A dataset of bees count data and relevant information in oilseed Brassica fields in an Australian temperate landscape.

Description

This dataset contains 212 samples of 61 variables including three bee species, inflorescence, temperature, wid speed and various derived landscape variables.

Usage

data("bees")

Format

A data frame with 212 observations on the following 61 variables.

transid

a factor with levels G1 G10 G11 G12 G13 G14 G15 G16 G1D1 G1D10 G1D11 G1D12 G1D13 G1D14 G1D15 G1D16 G1D17 G1D2 G1D3 G1D4 G1D5 G1D6 G1D7 G1D8 G1D9 G1G1 G1G10 G1G11 G1G12 G1G13 G1G14 G1G15 G1G16 G1G17 G1G2 G1G3 G1G4 G1G5 G1G6 G1G7 G1G8 G1G9 G2 G2H1 G2H10 G2H11 G2H2 G2H3 G2H4 G2H5 G2H6 G2H7 G2H8 G2H9 G2S1 G2S10 G2S11 G2S2 G2S3 G2S4 G2S5 G2S6 G2S7 G2S8 G2S9 G3 G4 G5 G6 G7 G8 G9 GCH1 GCH10 GCH11 GCH12 GCH13 GCH14 GCH15 GCH16 GCH17 GCH18 GCH2 GCH3 GCH4 GCH5 GCH6 GCH7 GCH8 GCH9 GCS1 GCS10 GCS11 GCS12 GCS13 GCS14 GCS15 GCS16 GCS17 GCS18 GCS2 GCS3 GCS4 GCS5 GCS6 GCS7 GCS8 GCS9 H1 H10 H11 H12 H13 H14 H15 H16 H2 H3 H4 H5 H6 H7 H8 H9 MC1G1 MC1G2 MC1G3 MC1G4 MC1G5 MC1G6 MC1G7 MC1G8 MC1G9 MC1H1 MC1H2 MC1H3 MC1H4 MC1H5 MC1H6 MC1H7 MC1H8 MC1H9 MC2AH1 MC2AH10 MC2AH11 MC2AH12 MC2AH13 MC2AH14 MC2AH15 MC2AH16 MC2AH17 MC2AH18 MC2AH2 MC2AH3 MC2AH4 MC2AH5 MC2AH6 MC2AH7 MC2AH8 MC2AH9 MC2AS1 MC2AS10 MC2AS11 MC2AS12 MC2AS13 MC2AS14 MC2AS15 MC2AS16 MC2AS17 MC2AS18 MC2AS2 MC2AS3 MC2AS4 MC2AS5 MC2AS6 MC2AS7 MC2AS8 MC2AS9 MC2BB1 MC2BB10 MC2BB11 MC2BB12 MC2BB13 MC2BB14 MC2BB15 MC2BB16 MC2BB17 MC2BB2 MC2BB3 MC2BB4 MC2BB5 MC2BB6 MC2BB7 MC2BB8 MC2BB9 MC2BG1 MC2BG10 MC2BG11 MC2BG12 MC2BG13 MC2BG14 MC2BG15 MC2BG16 MC2BG17 MC2BG2 MC2BG3 MC2BG4 MC2BG5 MC2BG6 MC2BG7 MC2BG8 MC2BG9

transsurv

a numeric vector

plotsurv

a numeric vector

paddock

a numeric vector

plot

a factor with levels G1-1 G1-10 G1-11 G1-12 G1-13 G1-14 G1-15 G1-16 G1-17 G1-2 G1-3 G1-4 G1-5 G1-6 G1-7 G1-8 G1-9 G21 G210 G211 G22 G23 G24 G25 G26 G27 G28 G29 GC1 GC10 GC11 GC12 GC13 GC14 GC15 GC16 GC17 GC18 GC2 GC3 GC4 GC5 GC6 GC7 GC8 GC9 MC-1 MC-10 MC-11 MC-12 MC-13 MC-14 MC-15 MC-16 MC-2 MC-3 MC-4 MC-5 MC-6 MC-7 MC-8 MC-9 MC1-1 MC1-2 MC1-3 MC1-4 MC1-5 MC1-6 MC1-7 MC1-8 MC1-9 MC2-A1 MC2-A10 MC2-A11 MC2-A12 MC2-A13 MC2-A14 MC2-A15 MC2-A16 MC2-A17 MC2-A18 MC2-A2 MC2-A3 MC2-A4 MC2-A5 MC2-A6 MC2-A7 MC2-A8 MC2-A9 MC2-B1 MC2-B10 MC2-B11 MC2-B12 MC2-B13 MC2-B14 MC2-B15 MC2-B16 MC2-B17 MC2-B2 MC2-B3 MC2-B4 MC2-B5 MC2-B6 MC2-B7 MC2-B8 MC2-B9

obs

a factor with levels Andrew Barbara Micah Sonia Steve

hbee

a numeric vector

nbee

a numeric vector

hover

a numeric vector

date

a numeric vector

sx

a numeric vector

fx

a numeric vector

sy

a numeric vector

fy

a numeric vector

loc

a factor with levels 100 150 200 400 450 edge

pair

a factor with levels I O

inf

a numeric vector

rankinf

a numeric vector

dupl

a numeric vector

temp

a numeric vector

windspeed

a numeric vector

winddir

a factor with levels N NE NNE NNW NW

cloudc

a numeric vector

disttoedgecalc

a numeric vector

disttoedgemeasured

a numeric vector

w100

a numeric vector

w200

a numeric vector

w300

a numeric vector

w400

a numeric vector

w500

a numeric vector

w600

a numeric vector

w700

a numeric vector

w800

a numeric vector

w900

a numeric vector

w1000

a numeric vector

w1500

a numeric vector

w2000

a numeric vector

c100

a numeric vector

c200

a numeric vector

c300

a numeric vector

c400

a numeric vector

c500

a numeric vector

c1000

a numeric vector

c1500

a numeric vector

area

a numeric vector

perimeter

a numeric vector

gyration

a numeric vector

paratio

a numeric vector

shape

a numeric vector

fractaldimention

a numeric vector

circumscircle

a numeric vector

contiguity

a numeric vector

links100

a numeric vector

links200

a numeric vector

links300

a numeric vector

links400

a numeric vector

links500

a numeric vector

links1000

a numeric vector

links1500

a numeric vector

links2000

a numeric vector

windspeed2

a numeric vector

Details

For details, please see the source. This dataset was published as an appendix of the paper listed in the source. Where the long and lat were reprojected to easting and northing

Source

The data source is [https://doi.org/10.25919/5f17b34638cca] or [https://data.csiro.au/collections/collection/CIcsiro:45533], which provides bees count data and relevant predictive variables along with a brief description of the data. The detailed descriptions of the data are available in: "Arthur, A. D., et al. (2010). "Influence of woody vegetation on pollinator densities in oilseed Brassica fields in an Australian temperate landscape." Basic and Applied Ecology 11: 406-414."

References

Arthur, A. D., Li, J., Henry, S., Cunningham, S.A. (2010). "Influence of woody vegetation on pollinator densities in oilseed Brassica fields in an Australian temperate landscape." Basic and Applied Ecology 11: 406-414.

Correct classification rate for predictive models based on cross -validation

Description

This function is to calculates correct classification rate (ccr) for categorical data with the observed (obs) data specified as factor. It based on the differences between the predicted values for and the observed values of validation samples for cross-validation. For 0 and 1 data, the observed values need to be specified as factor in order to use this accuracy measure. It is modified from the function 'pred.acc' in 'spm' package.

Usage

ccr(obs, pred)

Arguments

Value

A list with the following component: ccr (correct classification rate) for categorical data.

Author(s)

Jin Li

References

Jin Li (2019). spm: Spatial Predictive Modeling. R package version 1.2.0. https://CRAN.R-project.org/package=spm.

Examples

set.seed(1234)
x <- as.factor(sample(letters[1:2], 30, TRUE))
y <- sample(x, 30)
ccr(x, y)

Note on notes

Description

This is an updated and extended version of ‘spm' package. The change in package name from 'spm' to 'spm2' is due to the change in Author’s support from Geoscience Australia to Data2Action Australia.

## R CMD check results 0 errors | 0 warnings | 0 notes

Author(s)

Jin Li

Split data for k-fold cross-validation

Description

This function is a data splitting function for k-fold cross- validation and uses a stratified random sampling technique. It resamples the training data based on sample quantiles.

Usage

datasplit(trainy, k.fold = 10)

Arguments

Value

A list of samples each with an index of k-fold number.

Note

This function is largely based on rfcv in randomForest.

Author(s)

Jin Li

References

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Examples

library(spm)
data(petrel)
idx1 <- datasplit(petrel[, 3], k.fold = 10)
table(idx1)

Digit number after decimal point for a numeric variable

Description

This function is to derive the digit number after decimal point for a numeric variable (e.g., lat and long).

Usage

decimaldigit(x, dechar = ".", nint = NA, ndec = NA, pad.left = TRUE)

Arguments

Value

A list of integer number to show digit number after decimal point of x.

Note

This function is modified from decimal.align in 'prettyR' package.

Author(s)

Jin Li

References

Jim Lemon and Philippe Grosjean (2019). 'prettyR': Pretty Descriptive Stats. R package version 2.1.1. https://CRAN.R-project.org/package=prettyR.

Examples

x<-c(0.1, 2.2, 3.03, 44.444, 555.0005, 6666.66666)
decimaldigit(x)

Cross validation, n-fold and leave-one-out for the hybrid methods of generalized boosted regression modeling ('gbm'), 'kriging' and inverse distance weighted ('IDW').

Description

This function is a cross validation function for 38 hybrid methods of 'gbm', 'kriging' and 'IDW', including the average of 'gbmkrige' and 'gbmidw' ('gbmkrigegbmidw') and the average of 'gbm', 'gbmkrige' and 'gbmidw' ('gbmgbmkrigegbmidw'), where 'kriging' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') and 'IDW' also covers 'NN' and 'KNN'. The data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

gbmkrigeidwcv(
  longlat,
  trainx,
  trainy,
  var.monotone = rep(0, ncol(trainx)),
  family = "gaussian",
  n.trees = 3000,
  learning.rate = 0.001,
  interaction.depth = 2,
  bag.fraction = 0.5,
  train.fraction = 1,
  n.minobsinnode = 10,
  transformation = "none",
  weights = rep(1, nrow(trainx)),
  keep.data = FALSE,
  verbose = TRUE,
  delta = 1,
  formula = res1 ~ 1,
  vgm.args = "Sph",
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  n.cores = 6,
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'gbmcv' in 'spm', and 'krigecv' in 'spm2'.

Author(s)

Jin Li

References

Li, J. (2022). Spatial Predictive Modeling with R. Boca Raton, Chapman and Hall/CRC.

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

Greg Ridgeway with contributions from others (2015). gbm: Generalized Boosted Regression Models. R package version 2.1.1. https://CRAN.R-project.org/package=gbm

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
# gbmgbmokgbmidw
data(sponge)
longlat <- sponge[, 1:2]
set.seed(1234)
gbmgbmkrigegbmidwcv1 <- gbmkrigeidwcv(longlat = longlat,
trainx = sponge[, -3], trainy = sponge[, 3], family = "poisson", interaction.depth = 3,
transformation = "none", formula = res1 ~ 1, vgm.args = "Sph",
nmaxkrige = 12, idp = 2, nmaxidw = 12, hybrid.parameter = 3, validation = "CV",
predacc = "ALL", n.cores = 2)
gbmgbmkrigegbmidwcv1

# gbmokgbmidw for count data
data(sponge)
longlat <- sponge2[, 1:2]
y = sponge[, 3]
trainx = sponge[, -3]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
  gbmkrigegbmidwcv1 <- gbmkrigeidwcv(longlat = longlat,
  trainx = trainx, trainy = y, family = "poisson", interaction.depth = 3,
  transformation = "none", formula = res1 ~ 1, vgm.args = ("Sph"),
  nmaxkrige = 12, idp = 2, nmaxidw = 12, hybrid.parameter = 2, validation = "CV",
  predacc = "VEcv", n.cores = 2)
  VEcv [i] <- gbmkrigegbmidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for gbmokgbmidw", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid methods of generalized boosted regression modeling ('gbm'), 'kriging' and inverse distance weighted ('IDW').

Description

This function is for generating spatial predictions using the hybrid methods of 'gbm', 'kriging' and 'IDW', including all methods implemented in 'gbmkrigeidwcv'.

Usage

gbmkrigeidwpred(
  longlat,
  trainx,
  predx,
  trainy,
  longlatpredx,
  var.monotone = rep(0, ncol(trainx)),
  family = "gaussian",
  n.trees = 3000,
  learning.rate = 0.001,
  interaction.depth = 2,
  bag.fraction = 0.5,
  train.fraction = 1,
  n.minobsinnode = 10,
  transformation = "none",
  weights = rep(1, nrow(trainx)),
  keep.data = FALSE,
  verbose = TRUE,
  delta = 1,
  formula = res1 ~ 1,
  vgm.args = "Sph",
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  cv.fold = 10,
  n.cores = 8,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

Greg Ridgeway with contributions from others (2015). gbm: Generalized Boosted Regression Models. R package version 2.1.1. https://CRAN.R-project.org/package=gbm

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(sponge)
data(sponge.grid)
longlat <- sponge[, 1:2]

set.seed(1234)

gbmkrigeidwpred1 <- gbmkrigeidwpred(longlat = longlat, trainx = sponge[, -3],
predx = sponge.grid, trainy = sponge[, 3], longlatpredx = sponge.grid[, c(1:2)],
family = "poisson", interaction.depth = 3, transformation = "none", formula = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, hybrid.parameter = 3,
n.cores = 2)

names(gbmkrigeidwpred1)

range(gbmkrigeidwpred1$predictions)

Cross validation, n-fold and leave-one-out for generalised linear models ('glm')

Description

This function is a cross validation function for 'glm' method in 'stats' package.

Usage

glmcv(
  formula = NULL,
  trainxy,
  y,
  family = "gaussian",
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest' and 'glm' in 'stats'.

Author(s)

Jin Li

References

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Examples

library(spm)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
glmcv1 <- glmcv(formula = model, gravel, log(gravel[, 7] +1), validation = "CV",
 predacc = "ALL")
glmcv1 # Since the default 'family' is used, it is actually a 'lm' model.

data(sponge)
model <- sponge ~ easting + I(easting^2)
set.seed(1234)
glmcv1 <- glmcv(formula = model, sponge, sponge[, 3], family = poisson,
validation = "CV",  predacc = "ALL")
glmcv1

# For glm
model <- gravel / 100 ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glmcv1 <- glmcv(formula = model, gravel, gravel[, 7] / 100, family =
binomial(link=logit), validation = "CV", predacc = "VEcv")
VEcv [i] <- glmcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid method of generalised linear models ('glm') and inverse distance weighted ('IDW') ('glmidw')

Description

This function is a cross validation function for the hybrid method of 'glm' and 'idw' using 'gstat' (glmidw) (see reference #1), where the data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

glmidwcv(
  formula = NULL,
  longlat,
  trainxy,
  y,
  family = "gaussian",
  idp = 2,
  nmaxidw = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only.

Note

This function is largely based on 'rfcv' in 'randomForest', 'idwcv' in 'spm'and 'glm' in 'stats'.

Author(s)

Jin Li

References

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

set.seed(1234)
glmidwcv1 <- glmidwcv(formula = model, longlat = longlat, trainxy =  gravel,
y = y, idp = 2, nmaxidw = 12, validation = "CV", predacc = "ALL")
glmidwcv1 # Since the default 'family' is used, actually a 'lm' model is used.

data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ long + I(long^2)
y = spongelonglat[, 1]
set.seed(1234)
glmidwcv1 <- glmidwcv(formula = model, longlat = longlat, trainxy = spongelonglat,
y = y, family = poisson, idp = 2, nmaxidw = 12, validation = "CV",
predacc = "ALL")
glmidwcv1

# glmidw for count data
data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ . # use all predictive variables in the dataset
y = spongelonglat[, 1]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 glmidwcv1 <- glmidwcv(formula = model, longlat = longlat, trainxy = spongelonglat,
 y = y, family = poisson, idp = 2, nmaxidw = 12, validation = "CV",
 predacc = "VEcv")
 VEcv [i] <- glmidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

# glmidw for percentage data
longlat <- petrel[, c(1, 2)]
model <- gravel / 100 ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glmidwcv1 <- glmcv(formula = model, longlat = longlat, trainxy = gravel,
y = gravel[, 7] / 100, family = binomial(link=logit), idp = 2, nmaxidw = 12,
validation = "CV", predacc = "VEcv")
VEcv [i] <- glmidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid method of generalised linear models ('glm') and inverse distance weighted ('IDW') ('glmidw')

Description

This function is for generating spatial predictions using the hybrid method of 'glm' and 'idw' ('glmidw') (see reference #1).

Usage

glmidwpred(
  formula = NULL,
  longlat,
  trainxy,
  y,
  longlatpredx,
  predx,
  family = "gaussian",
  idp = 2,
  nmaxidw = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

glmidwpred1 <- glmidwpred(formula = model, longlat = longlat, trainxy =  gravel,
y = y, longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid, idp = 2,
 nmaxidw = 12)
 # Since the default 'family' is used, actually a 'lm' model is used.

names(glmidwpred1)

# Back transform 'glmidwpred$predictions' to generate the final predictions
glmidwpred1$predictions.bt <- exp(glmidwpred1$predictions) - 1
range(glmidwpred1$predictions.bt)

Cross validation, n-fold and leave-one-out for the hybrid method of generalised linear models ('glm') and 'krige' ('glmkrige')

Description

This function is a cross validation function for the hybrid method of 'glm' and 'krige' (glmkrige), where 'krige' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') (see reference #1 for further info).

Usage

glmkrigecv(
  formula.glm = NULL,
  longlat,
  trainxy,
  y,
  family = "gaussian",
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest', 'krigecv' in 'spm2'and 'glm' in 'stats'.

Author(s)

Jin Li

References

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)
set.seed(1234)
glmkrigecv1 <- glmkrigecv(formula.glm = model, longlat = longlat, trainxy =  gravel,
y = y, transformation = "none", formula.krige = res1 ~ 1, vgm.args = "Sph",
nmaxkrige = 12, validation = "CV", predacc = "ALL")
glmkrigecv1 # Since the default 'family' is used, actually a 'lm' model is used.

data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ long + I(long^2)
y = spongelonglat[, 1]
set.seed(1234)
glmkrigecv1 <- glmkrigecv(formula.glm = model, longlat = longlat, trainxy =
spongelonglat, y = y, family = poisson, transformation = "arcsine",
formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12,
validation = "CV", predacc = "ALL")
glmkrigecv1

# glmok for count data
data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ . # use all predictive variables in the dataset
y = spongelonglat[, 1]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 glmkrigecv1 <- glmkrigecv(formula.glm = model, longlat = longlat, trainxy = spongelonglat,
 y = y, family = poisson, formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12,
 validation = "CV",  predacc = "VEcv")
 VEcv [i] <- glmkrigecv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

# glmok for percentage data
longlat <- petrel[, c(1, 2)]
model <- gravel / 100 ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glmkrigecv1 <- glmkrigecv(formula.glm = model, longlat = longlat, trainxy = gravel,
y = gravel[, 7] / 100, family = binomial(link=logit), formula.krige = res1 ~ 1,
vgm.args = ("Sph"), nmaxkrige = 12, validation = "CV", predacc = "VEcv")
VEcv [i] <- glmkrigecv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid methods of generalised linear models ('glm'), 'kriging' and inverse distance weighted ('IDW').

Description

This function is a cross validation function for 38 hybrid methods of 'glm', 'kriging' and 'IDW', including the average of 'glmkrige' and 'glmidw' ('glmkrigeglmidw') and the average of 'glm', 'glmkrige' and 'glmidw' ('glmglmkrigeglmidw'), where 'kriging' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') and 'IDW' also covers 'NN' and 'KNN' (for details, see reference #1). This function can also be sued for 38 hybrid methods of 'lm', 'kriging' and 'IDW'.

Usage

glmkrigeidwcv(
  formula.glm = NULL,
  longlat,
  trainxy,
  y,
  family = "gaussian",
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest', 'krigecv' in 'spm2'and 'glm' in 'stats'.

Author(s)

Jin Li

References

Li, J. (2022). Spatial Predictive Modeling with R. Boca Raton, Chapman and Hall/CRC.

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
# glmokglidw
data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)
set.seed(1234)
glmkrigeglmidwcv1 <- glmkrigeidwcv(formula.glm = model, longlat = longlat,
trainxy =  gravel, y = y, transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",
 predacc = "ALL")
glmkrigeglmidwcv1 # Since the default 'family' is used, actually a 'lm' model is used.

# glmokglmidw
data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ long + I(long^2)
y = spongelonglat[, 1]
set.seed(1234)
glmkrigeglmidwcv1 <- glmkrigeidwcv(formula.glm = model, longlat = longlat,
trainxy = spongelonglat, y = y, family = poisson, transformation = "arcsine",
formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12, idp = 2,
nmaxidw = 12, validation = "CV", predacc = "ALL")
glmkrigeglmidwcv1

# glmglmokglmidw
data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ long + I(long^2)
y = spongelonglat[, 1]
set.seed(1234)
glmglmkrigeglmidwcv1 <- glmkrigeidwcv(formula.glm = model, longlat = longlat,
trainxy = spongelonglat, y = y, family = poisson, transformation = "arcsine",
formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12, idp = 2,
nmaxidw = 12, hybrid.parameter = 3, validation = "CV", predacc = "ALL")
glmglmkrigeglmidwcv1

# glmokglidw for count data
data(spongelonglat)
longlat <- spongelonglat[, 7:8]
model <- sponge ~ . # use all predictive variables in the dataset
y = spongelonglat[, 1]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 glmkrigeglmidwcv1 <- glmkrigeidwcv(formula.glm = model, longlat = longlat,
 trainxy = spongelonglat, y = y, family = poisson, formula.krige = res1 ~ 1,
 vgm.args = ("Sph"), nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",
 predacc = "VEcv")
 VEcv [i] <- glmkrigeglmidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

# glmokglmidw for percentage data
longlat <- petrel[, c(1, 2)]
model <- gravel / 100 ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glmkrigeglmidwcv1 <- glmkrigeidwcv(formula.glm = model, longlat = longlat,
trainxy = gravel, y = gravel[, 7] / 100, family = binomial(link=logit),
formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12, idp = 2,
nmaxidw = 12, validation = "CV", predacc = "VEcv")
VEcv [i] <- glmkrigeglmidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLM", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid methods of generalised linear models ('glm'), 'kriging' and inverse distance weighted ('IDW').

Description

This function is for generating spatial predictions using the hybrid methods of 'glm', 'kriging' and 'IDW', including all methods implemented in 'glmkrigeidwcv'.

Usage

glmkrigeidwpred(
  formula.glm = NULL,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  family = "gaussian",
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

glmkrigeidwpred1 <- glmkrigeidwpred(formula.glm = model, longlat = longlat, trainxy =  gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12)
 # Since the default 'family' is used, actually a 'lm' model is used.

names(glmkrigeidwpred1)

# Back transform 'glmkrigeidwpred$predictions' to generate the final predictions
glmkrigeidw.predictions <- exp(glmkrigeidwpred1$predictions) - 1
range(glmkrigeidw.predictions)

Generate spatial predictions using the hybrid method of generalised linear models ('glm') and 'krige'

Description

This function is for generating spatial predictions using the hybrid method of 'glm' and 'krige', including all methods implemented in 'glmkrigecv'. (see reference #1 for further info).

Usage

glmkrigepred(
  formula.glm = NULL,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  family = "gaussian",
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Alvarez, B., Siwabessy, J., Tran, M., Huang, Z., Przeslawski, R., Radke, L., Howard, F. and Nichol, S. (2017). "Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness." Environmental Modelling & Software 97: 112-129.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

glmkrigepred1 <- glmkrigepred(formula.glm = model, longlat = longlat, trainxy =  gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
transformation = "none", formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12)
 # Since the default 'family' is used, actually a 'lm' model is used.

names(glmkrigepred1)

# Back transform 'glmkrigepred$predictions' to generate the final predictions
glmkrige.predictions <- exp(glmkrigepred1$predictions) - 1
range(glmkrige.predictions)

Cross validation, n-fold and leave-one-out, for 'glmnet' in 'glmnet' package

Description

This function is a cross validation function for 'glmnet' method in 'glmnet' package.

Usage

glmnetcv(
  trainx,
  y,
  family = "gaussian",
  alpha = 0.5,
  relax = FALSE,
  type.measure = "mse",
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'glmcv' in this 'spm2' package.

Author(s)

Jin Li

References

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Examples

library(spm)

data(petrel)
x <- as.matrix(petrel[, c(1, 2, 6:9)])
y <- log(petrel[, 5] + 1)
set.seed(1234)
glmnetcv1 <- glmnetcv(x, y, validation = "CV",  predacc = "ALL")
glmnetcv1

data(sponge)
x <- as.matrix(cbind(sponge$easting, sponge$easting^2))
set.seed(1234)
glmnetcv1 <- glmnetcv(x, sponge[, 3], family = poisson, validation = "CV",
predacc = "ALL")
glmnetcv1

# For glmnet with gaussian
x <- as.matrix(petrel[, c(1, 2, 6:9)])
y <- log(petrel[, 5] + 1)
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
  glmnetcv1 <- glmnetcv(x, y, validation = "CV", predacc = "VEcv")
  VEcv [i] <- glmnetcv1
 }
plot(VEcv ~ c(1:n), xlab = "Iteration for glmnet", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

# For glmnet with binomial
x <- as.matrix(cbind(petrel[, c(2, 6)], petrel$long^3, petrel$lat^2, petrel$lat^3))
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glmnetcv1 <- glmnetcv(x, petrel[, 5] / 100, family = binomial(link=logit),
validation = "CV", predacc = "VEcv")
VEcv [i] <- glmnetcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for glmnet", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using generalised linear models ('glm')

Description

This function is for generating spatial predictions using 'glm' method in 'stats' package.

Usage

glmpred(formula = NULL, trainxy, longlatpredx, predx, family = "gaussian", ...)

Arguments

Value

A dataframe of longitude, latitude and predictions.

Author(s)

Jin Li

Examples

library(spm)
data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)

glmpred1 <- glmpred(formula = model, trainxy = gravel,
longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid)

names(glmpred1)

# Back transform 'glmpred1$pred.glm1' to generate the final predictions
glm.predictions <- exp(glmpred1$pred.glm1) - 1
range(glm.predictions)

Cross validation, n-fold and leave-one-out for generalized least squares ('gls')

Description

This function is a cross validation function for 'gls' method in 'nlme' package.

Usage

glscv(
  model = var1 ~ 1,
  trainxy,
  y,
  corr.args = NULL,
  weights = NULL,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on rfcv in 'randomForest' and 'gls' in 'library(nlme)'.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Examples

library(spm)
library(nlme)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
range1 <- 0.8
nugget1 <- 0.5

model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)


glscv1 <- glscv(model = model, gravel, log(gravel[, 7] +1), validation = "CV",
 corr.args = corSpher(c(range1, nugget1), form = ~ long + lat, nugget = TRUE),
 predacc = "ALL")
glscv1

#For gls
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glscv1 <- glscv(model = model, gravel, log(gravel[, 7] +1), validation = "CV",
          corr.args = corSpher(c(range1, nugget1), form = ~ long + lat,
          nugget = TRUE), predacc = "VEcv")
VEcv [i] <- glscv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLS", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

# For lm, that is, gls with 'correlation = NULL'
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
set.seed(1234)
for (i in 1:n) {
glscv1 <- glscv(model = model, gravel, log(gravel[, 7] +1),
validation = "CV", predacc = "VEcv")
VEcv [i] <- glscv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLS", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid method of generalized least squares ('gls') and inverse distance weighted ('idw') (glsidw)

Description

This function is a cross validation function for the hybrid method of 'gls' and 'idw', where the data splitting is based on a stratified random sampling method (see the 'datasplit' function for details)

Usage

glsidwcv(
  model = var1 ~ 1,
  longlat,
  trainxy,
  y,
  corr.args = NULL,
  weights = NULL,
  idp = 2,
  nmaxidw = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only.

Note

This function is largely based on rfcv in 'randomForest' and 'gls' in 'library(nlme)'.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)

glsidwcv1 <- glsidwcv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), idp = 2, nmaxidw = 12, validation = "CV",
 corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
 predacc = "ALL")
glsidwcv1

# For glsidw
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glsidwcv1 <- glsidwcv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), idp = 2, nmaxidw = 12, validation = "CV",
corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
predacc = "VEcv")
VEcv [i] <- glsidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLSIDW", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid method of generalized least squares ('gls') and inverse distance weighted ('IDW') ('glsidw')

Description

This function is for generating spatial predictions using the hybrid method of 'gls' and 'idw' ('glsidw') (see reference #1).

Usage

glsidwpred(
  model = var1 ~ 1,
  longlat,
  trainxy,
  y,
  longlatpredx,
  predx,
  corr.args = NULL,
  weights = NULL,
  idp = 2,
  nmaxidw = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)
y <- log(gravel[, 7] +1)

glsidwpred1 <- glsidwpred(model = model, longlat = longlat, trainxy = gravel,
y = y, longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid,
 idp = 2, nmaxidw = 12, corr.args = corSpher(c(range1, nugget1),
 form = ~ lat + long, nugget = TRUE))

names(glsidwpred1)

# Back transform 'glsidwpred$predictions' to generate the final predictions
glsidw.predictions <- exp(glsidwpred1$predictions) - 1
range(glsidw.predictions)

Cross validation, n-fold and leave-one-out for the hybrid method of generalized least squares ('gls') and kriging ('krige') ('glskrige')

Description

This function is a cross validation function for the hybrid method of 'gls' and 'krige' ('glskrige'), where the data splitting is based on a stratified random sampling method (see the 'datasplit' function for details)

Usage

glskrigecv(
  model = var1 ~ 1,
  longlat,
  trainxy,
  y,
  corr.args = NULL,
  weights = NULL,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only.

Note

This function is largely based on rfcv in 'randomForest', 'krigecv' in 'spm2' and 'gls' in 'library(nlme)'.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)

glskrigecv1 <- glskrigecv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, validation = "CV",
 corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
 predacc = "ALL")
glskrigecv1

# For glskrige
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glskrigecv1 <- glskrigecv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxok = 12, validation = "CV",
corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
 predacc = "VEcv")
VEcv [i] <- glskrigecv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLSOK", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid methods of generalised least squares ('gls'), 'kriging' and inverse distance weighted ('IDW')

Description

This function is a cross validation function for 38 hybrid methods of 'gls', 'kriging' and 'IDW', including the average of 'glskrige' and 'glsidw' ('glskrigeglsidw') and the average of 'gls', 'glskrige' and 'glsidw' ('glsglskrigeglsidw'), where 'kriging' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') and 'IDW' also covers 'NN' and 'KNN'.. The data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

glskrigeidwcv(
  model = var1 ~ 1,
  longlat,
  trainxy,
  y,
  corr.args = NULL,
  weights = NULL,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only.

Note

This function is largely based on rfcv in 'randomForest', 'krigecv' in 'spm2' and 'gls' in 'library(nlme)'.

Author(s)

Jin Li

References

Li, J. (2022). Spatial Predictive Modeling with R. Boca Raton, Chapman and Hall/CRC.

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)

glskrigeidwcv1 <- glskrigeidwcv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",
 corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
 predacc = "ALL")
glskrigeidwcv1

# For glskrigeglsidw
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glskrigeidwcv1 <- glskrigeidwcv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",
corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE),
 predacc = "VEcv")
VEcv [i] <- glskrigeidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLSOKGLSIDW", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)


# For glsglskrigeglsidw
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
glskrigeidwcv1 <- glskrigeidwcv(model = model, longlat = longlat, trainxy = gravel,
y = log(gravel[, 7] +1), transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, hybrid.parameter = 3,
validation = "CV", corr.args = corSpher(c(range1, nugget1), form = ~ lat + long,
 nugget = TRUE), predacc = "VEcv")
VEcv [i] <- glskrigeidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for GLSOKGLSIDW", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid methods of generalised least squares ('gls'), 'kriging' and inverse distance weighted ('IDW')

Description

This function is for generating spatial predictions using the hybrid methods of 'gls', 'kriging' and 'IDW', including all methods implemented in 'glskrigeidwcv'.

Usage

glskrigeidwpred(
  model = var1 ~ 1,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  corr.args = NULL,
  weights = NULL,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)
y <- log(gravel[, 7] +1)

glskrigeidwpred1 <- glskrigeidwpred(model = model, longlat = longlat, trainxy = gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
transformation = "none", formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12,
idp = 2, nmaxidw = 12,  corr.args = corSpher(c(range1, nugget1),
form = ~ lat + long, nugget = TRUE))

names(glskrigeidwpred1)

# Back transform 'glskrigeidwpred$predictions' to generate the final predictions
glskrigeidw.predictions <- exp(glskrigeidwpred1$predictions) - 1
range(glskrigeidw.predictions)

Generate spatial predictions using the hybrid method of generalized least squares ('gls') and kriging ('krige') ('glskrige')

Description

This function is for generating spatial predictions using the hybrid method of 'gls' and 'krige' (glskrige).

Usage

glskrigepred(
  model = var1 ~ 1,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  corr.args = NULL,
  weights = NULL,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
library(nlme)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
range1 <- 0.8
nugget1 <- 0.5
model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)
y <- log(gravel[, 7] +1)

glskrigepred1 <- glskrigepred(model = model, longlat = longlat, trainxy = gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, corr.args = corSpher(c(range1, nugget1),
form = ~ lat + long, nugget = TRUE))

names(glskrigepred1)

# Back transform 'glskrigepred$predictions' to generate the final predictions
glskrige.predictions <- exp(glskrigepred1$predictions) - 1
range(glskrige.predictions)

Generate spatial predictions using generalized least squares ('gls')

Description

This function is for generating spatial predictions using 'gls' method in 'nlme' package.

Usage

glspred(
  model = var1 ~ 1,
  trainxy,
  longlatpredx,
  predx,
  corr.args = NULL,
  weights = NULL,
  ...
)

Arguments

Value

A dataframe of longitude, latitude and predictions.

Author(s)

Jin Li

References

Pinheiro, J. C. and D. M. Bates (2000). Mixed-Effects Models in S and S-PLUS. New York, Springer.

Examples

library(spm)
library(nlme)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
range1 <- 0.8
nugget1 <- 0.5

model <- log(gravel + 1) ~  long + lat +  bathy + dist + I(long^2) + I(lat^2) +
I(lat^3) + I(bathy^2) + I(bathy^3) + I(dist^2) + I(dist^3) + I(relief^2) + I(relief^3)

glspred1 <- glspred(model = model, trainxy = gravel,
 longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid,
 corr.args = corSpher(c(range1, nugget1), form = ~ lat + long, nugget = TRUE))

names(glspred1)

# Back transform 'glspred1$predictions' to generate the final predictions
gls.predictions <- exp(glspred1$predictions) - 1
range(gls.predictions)

Cross validation, n-fold and leave-one-out for kriging methods ('krige')

Description

This function is a cross validation function for kriging methods ('krige') in 'gstat'.

Usage

krigecv(
  longlat,
  trainy,
  trainpredx = NULL,
  validation = "CV",
  cv.fold = 10,
  nmax = 12,
  transformation = "none",
  delta = 1,
  formula = var1 ~ 1,
  vgm.args = ("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on rfcv in 'randomForest' and some functions in 'library(gstat)'. When 'A zero or negative range was fitted to variogram' occurs, to allow 'gstat' running, the range was set to be positive by using 'min(vgm1$dist)'. In this case, caution should be taken in applying this method. If it still occurs for 'okpred' function, different method may need to be used.

Author(s)

Jin Li

References

Li, J., 2013. Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques in Marine Environmental Geosciences, In: Christen, P., Kennedy, P., Liu, L., Ong, K.-L., Stranieri, A., Zhao, Y. (Eds.), The proceedings of the Eleventh Australasian Data Mining Conference (AusDM 2013), Canberra, Australia, 13-15 November 2013. Conferences in Research and Practice in Information Technology, Vol. 146.

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(sp)
library(spm)
data(swmud)
data(petrel)

set.seed(1234)
okcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", vgm.args = ("Sph"), predacc = "VEcv")
okcv1

set.seed(1234)
skcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", vgm.args = ("Sph"), predacc = "VEcv", beta = mean(swmud[, 3]))
skcv1

set.seed(1234)
ukcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", formula = var1 ~ long + lat, vgm.args = ("Sph"), predacc = "VEcv")
ukcv1

set.seed(1234)
okcv2 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], validation = "LOO", nmax = 7,
transformation = "arcsine", vgm.args = ("Sph"), predacc = "ALL")
okcv2

set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
okcv1 <- krigecv(longlat = petrel[, c(1,2)], trainy = petrel[, 5], nmax = 12,
transformation = "arcsine", predacc = "VEcv")
VEcv [i] <- okcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for OK", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

set.seed(1234)
n <- 20 # number of iterations, 60 to 100 is recommended.
measures <- NULL
for (i in 1:n) {
okcv1 <- krigecv(longlat = petrel[, c(1,2)], trainy = petrel[, 3], nmax = 12, transformation =
"arcsine", predacc = "ALL")
measures <- rbind(measures, okcv1$vecv)
}
plot(measures ~ c(1:n), xlab = "Iteration for OK", ylab = "VEcv (%)")
points(cumsum(measures) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(measures), col = 'blue', lwd = 2)

Generate spatial predictions using kriging methods ('krige')

Description

This function is to make spatial predictions using kriging methods ('krige').

Usage

krigepred(
  trainx,
  trainy,
  trainx2,
  nmax = 12,
  transformation = "none",
  delta = 1,
  formula = var1 ~ 1,
  vgm.args = ("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, predictions and variances.

Note

The variances in the output are not transformed back when a transformation is used. This is because kriging variances are not measuring the uncertainty of predictions but they are indicator of the spatial distribution of sample density. The variances are exported only for interested users; and if needed, they can be transformed back from the output.

Author(s)

Jin Li

References

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(sp)
library(spm)
data(swmud)
data(sw)
okpred1 <- krigepred(swmud[, c(1,2)], swmud[, 3], sw, nmax = 7, transformation =
"arcsine", vgm.args = ("Sph"))
names(okpred1)

Cross validation, n-fold and leave-one-out for the hybrid methods of 'random forest' ('RF'), 'kriging' and inverse distance weighted ('IDW')

Description

This function is a cross validation function for 38 hybrid methods of 'RF', 'kriging' and 'IDW', including the average of 'rfkrige' and 'rfidw' ('rfkrigerfidw') and the average of 'rf', 'rfkrige' and 'rfidw' ('rfrfkrigerfidw'), where 'kriging' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') and 'IDW' also covers 'NN' and 'KNN'.. The data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

rfkrigeidwcv(
  longlat,
  trainx,
  trainy,
  mtry = function(p) max(1, floor(sqrt(p))),
  ntree = 500,
  transformation = "none",
  delta = 1,
  formula = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest', and 'krigecv' in 'spm2'.

Author(s)

Jin Li

References

Li, J. (2022). Spatial Predictive Modeling with R. Boca Raton, Chapman and Hall/CRC.

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

Liaw, A. and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
# rfrfokrfidw
data(sponge)
longlat <- sponge[, 1:2]
set.seed(1234)
rfrfkrigerfidwcv1 <- rfkrigeidwcv(longlat = longlat,
trainx = sponge[, -3], trainy = sponge[, 3], formula = res1 ~ 1, vgm.args = ("Sph"),
nmaxkrige = 12, idp = 2, nmaxidw = 12, hybrid.parameter = 3, validation = "CV",
predacc = "ALL")
rfrfkrigerfidwcv1

# rfokrfidw for count data
data(sponge)
longlat <- sponge2[, 1:2]
y = sponge[, 3]
trainx = sponge[, -3]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 rfkrigerfidwcv1 <- rfkrigeidwcv(longlat = longlat,
 trainx = trainx, trainy = y, formula = res1 ~ 1, vgm.args = ("Sph"),
 nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",  predacc = "VEcv")
 VEcv [i] <- rfkrigerfidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for rfokrfidw", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid methods of 'random forest' ('RF'), 'kriging' and inverse distance weighted ('IDW').

Description

This function is for generating spatial predictions using the hybrid methods of 'RF', 'kriging' and 'IDW', including all methods implemented in 'rfkrigeidwcv'.

Usage

rfkrigeidwpred(
  longlat,
  trainx,
  predx,
  trainy,
  longlatpredx,
  mtry = function(p) max(1, floor(sqrt(p))),
  ntree = 500,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

Liaw, A. and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(sponge)
data(sponge.grid)
longlat <- sponge[, 1:2]
y = sponge[, 3]
trainx = sponge[, -3]

set.seed(1234)

rfkrigeidwpred1 <- rfkrigeidwpred(longlat = longlat, trainx =  trainx,
predx = sponge.grid, trainy = y, longlatpredx = sponge.grid[, c(1:2)],
formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12)

names(rfkrigeidwpred1)

range(rfkrigeidwpred1$predictions)

A dataset of sponge species richness in the Timor Sea region, northern Australia marine margin

Description

This dataset contains 77 samples of 81 variables including easting (longitude), northing (latitude), bathy, backscatter and their derived variables.

Usage

data("sponge2")

Format

A data frame with 77 observations on the following 89 variables.

easting

a numeric vector, m

northing

a numeric vector, m

species.richness

a numeric vector, no unit

mud

a numeric vector, percentage

sand

a numeric vector, percentage

gravel

a numeric vector, percentage

bathy

a numeric vector, m

bs25

a numeric vector, dB

bs10

a numeric vector, dB

bs11

a numeric vector, dB

bs12

a numeric vector, dB

bs13

a numeric vector, dB

bs14

a numeric vector, dB

bs15

a numeric vector, dB

bs16

a numeric vector, dB

bs17

a numeric vector, dB

bs18

a numeric vector, dB

bs19

a numeric vector, dB

bs20

a numeric vector, dB

bs21

a numeric vector, dB

bs22

a numeric vector, dB

bs23

a numeric vector, dB

bs24

a numeric vector, dB

bs26

a numeric vector, dB

bs27

a numeric vector, dB

bs28

a numeric vector, dB

bs29

a numeric vector, dB

bs30

a numeric vector, dB

bs31

a numeric vector, dB

bs32

a numeric vector, dB

bs33

a numeric vector, dB

bs34

a numeric vector, dB

bs35

a numeric vector, dB

bs36

a numeric vector, dB

bs_o

a numeric vector, dB

bs_homo_o

a numeric vector

bs_entro_o

a numeric vector, no unit

bs_var_o

a numeric vector, dB^2

bs_lmi_o

a numeric vector

bathy_o

a numeric vector, m

bathy_lmi_o

a numeric vector

tpi_o

a numeric vector, no unit

slope_o

a numeric vector

plan_cur_o

a numeric vector

prof_cur_o

a numeric vector

relief_o

a numeric vector

rugosity_o

a numeric vector

dist.coast

a numeric vector, m

rugosity3

a numeric vector

rugosity5

a numeric vector

rugosity7

a numeric vector

tpi3

a numeric vector, no unit

tpi5

a numeric vector, no unit

tpi7

a numeric vector, no unit

bathy_lmi3

a numeric vector

bathy_lmi5

a numeric vector

bathy_lmi7

a numeric vector

plan_curv3

a numeric vector

plan_curv5

a numeric vector

plan_curv7

a numeric vector

relief_3

a numeric vector

relief_5

a numeric vector

relief_7

a numeric vector

slope3

a numeric vector

slope5

a numeric vector

slope7

a numeric vector

prof_cur3

a numeric vector

prof_cur5

a numeric vector

prof_cur7

a numeric vector

entro3

a numeric vector, no unit

entro5

a numeric vector, no unit

entro7

a numeric vector, no unit

homo3

a numeric vector

homo5

a numeric vector

homo7

a numeric vector

var3

a numeric vector, dB^2

var5

a numeric vector, dB^2

var7

a numeric vector, dB^2

bs_lmi3

a numeric vector

bs_lmi5

a numeric vector

bs_lmi7

a numeric vector

Details

For details, please see the source. This dataset was published as an appendix of the paper listed in the source. Where the long and lat were reprojected to easting and northing.

Source

see Appendix A-D. Supplementary data at: "http://dx.doi.org/10.1016/j.envsoft.2017.07.016."

References

Li, J., B. Alvarez, J. Siwabessy, M. Tran, Z. Huang, R. Przeslawski, L. Radke, F. Howard, and S. Nichol. 2017. Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness. Environmental Modelling & Software, 97: 112-129.

A dataset of sponge species richness in the Timor Sea region, northern Australia marine margin

Description

This dataset contains 77 samples of 7 predictive variables including longitude, latitude, bathy, backscatter and their derived variables. It is the sponge dataset in 'spm' package, but with long and lat instead of easting and northing.

Usage

data("spongelonglat")

Format

A data frame with 77 observations on the following 8 variables.

sponge

a numeric vector

tpi3

a numeric vector

var7

a numeric vector

entro7

a numeric vector

bs34

a numeric vector

bs11

a numeric vector

long

a numeric vector

lat

a numeric vector

Details

For details, please see sponge dataset in library(spm). Where the long and lat were projected to easting and northing.

Source

sponge dataset in library(spm)

References

Li, J., B. Alvarez, J. Siwabessy, M. Tran, Z. Huang, R. Przeslawski, L. Radke, F. Howard, and S. Nichol. 2017. Application of random forest, generalised linear model and their hybrid methods with geostatistical techniques to count data: Predicting sponge species richness. Environmental Modelling & Software, 97: 112-129.

Examples

data(spongelonglat)
## maybe str(spongelonglat) ; plot(spongelonglat) ...

Cross validation, n-fold and leave-one-out for support vector machine ('svm')

Description

This function is a cross validation function for 'svm' regression in 'e1071' package.

Usage

svmcv(
  formula = NULL,
  trainxy,
  y,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest' and 'svm' in 'e1071'.

Author(s)

Jin Li

References

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Examples

library(spm)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
model <- log(gravel + 1) ~ lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
set.seed(1234)
svmcv1 <- svmcv(formula = model, gravel, log(gravel[, 7] +1), validation = "CV",
 predacc = "ALL")
svmcv1

data(sponge2)
model <- species.richness ~ .
set.seed(1234)
svmcv1 <- svmcv(formula = model, sponge2[, -4], sponge[, 3], gamma = 0.01, cost = 3.5,
scale = TRUE, validation = "CV",  predacc = "VEcv")
svmcv1

# For svm
model <- species.richness ~ .
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
svmcv1 <- svmcv(formula = model, sponge2[, -4], sponge[, 3], gamma = 0.01, cost = 3.5,
scale = TRUE, validation = "CV",  predacc = "VEcv")
VEcv [i] <- svmcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for svm", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid method of support vector machine ('svm') regression and inverse distance weighted ('IDW') (svmidw)

Description

This function is a cross validation function for the hybrid method of 'svm' regression and 'idw' using 'gstat' (svmidw), where the data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

svmidwcv(
  formula = NULL,
  longlat,
  trainxy,
  y,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  idp = 2,
  nmaxidw = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only.

Note

This function is largely based on 'rfcv' in 'randomForest', 'idwcv' in 'spm'and 'svm' in 'e1071'.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

set.seed(1234)
svmidwcv1 <- svmidwcv(formula = model, longlat = longlat, trainxy =  gravel,
y = y, idp = 2, nmaxidw = 12, validation = "CV", predacc = "ALL")
svmidwcv1

# svmidw for count data
data(sponge2)
model <- species.richness ~ . # use all predictive variables in the dataset
longlat <- sponge2[, 1:2]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 svmidwcv1 <- svmidwcv(formula = model, longlat = longlat, trainxy = sponge2[, -4],
 y = sponge[, 3], gamma = 0.01,  cost = 3.5, scale = TRUE, idp = 2, nmaxidw = 12,
 validation = "CV", predacc = "VEcv")
 VEcv [i] <- svmidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for svmidw", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid method of support vector machine ('svm') regression and inverse distance weighted ('IDW') ('svmidw')

Description

This function is for generating spatial predictions using the hybrid method of 'svm' and 'idw' ('svmidw').

Usage

svmidwpred(
  formula = NULL,
  longlat,
  trainxy,
  y,
  longlatpredx,
  predx,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  idp = 2,
  nmaxidw = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

svmidwpred1 <- svmidwpred(formula = model, longlat = longlat, trainxy =  gravel,
y = y, longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid, idp = 2,
 nmaxidw = 12)

names(svmidwpred1)

# Back transform 'svmidwpred$predictions' to generate the final predictions
svmidw.predictions <- exp(svmidwpred1$predictions) - 1
range(svmidw.predictions)

Cross validation, n-fold and leave-one-out for the hybrid method of support vector machine ('svm') regression and 'krige' (svmkrige)

Description

This function is a cross validation function for the hybrid method of 'svm' regression and 'krige' (svmkrige), where the data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

svmkrigecv(
  formula.svm = NULL,
  longlat,
  trainxy,
  y,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest', 'krigecv' in 'spm2'and 'svm' in 'e1071'.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)
set.seed(1234)
svmkrigecv1 <- svmkrigecv(formula.svm = model, longlat = longlat, trainxy =  gravel,
y = y, transformation = "none", formula.krige = res1 ~ 1, vgm.args = "Sph",
nmaxkrige = 12, validation = "CV", predacc = "ALL")
svmkrigecv1

# svmok for count data
data(sponge2)
model <- species.richness ~ . # use all predictive variables in the dataset
longlat <- sponge2[, 1:2]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 svmkrigecv1 <- svmkrigecv(formula.svm = model, longlat = longlat, trainxy = sponge2[, -4],
 y = sponge2[, 3], gamma = 0.01, cost = 3.5, scale = TRUE,  formula.krige = res1 ~ 1,
 vgm.args = ("Sph"), nmaxkrige = 12,  validation = "CV",  predacc = "VEcv")
 VEcv [i] <- svmkrigecv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for svm", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

Cross validation, n-fold and leave-one-out for the hybrid methods of support vector machine ('svm') regression , 'kriging' and inverse distance weighted ('IDW').

Description

This function is a cross validation function for 38 hybrid methods of 'svm', 'kriging' and 'IDW', including the average of 'svmkrige' and 'svmidw' ('svmkrigesvmidw') and the average of 'svm', 'svmkrige' and 'svmidw' ('svmsvmkrigesvmidw'), where 'kriging' methods include ordinary kriging ('OK'), simple kriging ('SK'), block 'OK' ('BOK') and block 'SK'('BSK') and 'IDW' also covers 'NN' and 'KNN'.. The data splitting is based on a stratified random sampling method (see the 'datasplit' function for details).

Usage

svmkrigeidwcv(
  formula.svm = NULL,
  longlat,
  trainxy,
  y,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'rfcv' in 'randomForest', 'krigecv' in 'spm2'and 'svm' in 'e1071'.

Author(s)

Jin Li

References

Li, J. (2022). Spatial Predictive Modeling with R. Boca Raton, Chapman and Hall/CRC.

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)
# svmokglidw
data(petrel)
gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)
set.seed(1234)
svmkrigesvmidwcv1 <- svmkrigeidwcv(formula.svm = model, longlat = longlat,
trainxy =  gravel, y = y, transformation = "none", formula.krige = res1 ~ 1,
vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12, validation = "CV",
 predacc = "ALL")
svmkrigesvmidwcv1

# svmsvmoksvmidw
data(sponge2)
model <- species.richness ~ . # use all predictive variables in the dataset
longlat <- sponge2[, 1:2]
y = sponge[, 3]
set.seed(1234)
svmsvmkrigesvmidwcv1 <- svmkrigeidwcv(formula.svm = model, longlat = longlat,
trainxy = sponge2[, -4], y = y, gamma = 0.01, cost = 3.5, scale = TRUE,
formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12, idp = 2,
nmaxidw = 12, hybrid.parameter = 3, validation = "CV", predacc = "ALL")
svmsvmkrigesvmidwcv1

# svmoksvmidw for count data
data(sponge2)
model <- species.richness ~ . # use all predictive variables in the dataset
longlat <- sponge2[, 1:2]
y = sponge2[, 3]
set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
 svmkrigesvmidwcv1 <- svmkrigeidwcv(formula.svm = model, longlat = longlat,
 trainxy = sponge2[, -4], y = y, gamma = 0.01, cost = 3.5, scale = TRUE,
 formula.krige = res1 ~ 1, vgm.args = ("Sph"), nmaxkrige = 12, idp = 2,
 nmaxidw = 12, validation = "CV",  predacc = "VEcv")
 VEcv [i] <- svmkrigesvmidwcv1
 }
 plot(VEcv ~ c(1:n), xlab = "Iteration for svm", ylab = "VEcv (%)")
 points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
 abline(h = mean(VEcv), col = 'blue', lwd = 2)

Generate spatial predictions using the hybrid methods of support vector machine ('svm') regression , 'kriging' and inverse distance weighted ('IDW').

Description

This function is for generating spatial predictions using the hybrid methods of 'svm', 'kriging' and 'IDW', including all methods implemented in 'svmkrigeidwcv'.

Usage

svmkrigeidwpred(
  formula.svm = NULL,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  idp = 2,
  nmaxidw = 12,
  hybrid.parameter = 2,
  lambda = 1,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

svmkrigeidwpred1 <- svmkrigeidwpred(formula.svm = model, longlat = longlat, trainxy =  gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12, idp = 2, nmaxidw = 12)

names(svmkrigeidwpred1)

# Back transform 'svmkrigeidwpred$predictions' to generate the final predictions
svmkrigeidw.predictions <- exp(svmkrigeidwpred1$predictions) - 1
range(svmkrigeidw.predictions)

Generate spatial predictions using the hybrid method of support vector machine ('svm') regression and 'krige' (svmkrige)

Description

This function is for generating spatial predictions using the hybrid method of 'svm' and 'krige' (svmkrige).

Usage

svmkrigepred(
  formula.svm = NULL,
  longlat,
  trainxy,
  predx,
  y,
  longlatpredx,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  transformation = "none",
  delta = 1,
  formula.krige = res1 ~ 1,
  vgm.args = c("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  nmaxkrige = 12,
  ...
)

Arguments

Value

A dataframe of longitude, latitude, and predictions.

Author(s)

Jin Li

References

Li, J., Potter, A., Huang, Z., and Heap, A. (2012). Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and Geostatistical Methods, Geoscience Australia, Record 2012/48, 115pp.

Li, J., Heap, A., Potter, A., and Danilel, J.J. (2011). Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine Learning Methods and Their Combination with Ordinary Kriging and Inverse Distance Squared, Geoscience Australia, Record 2011/07, 69pp.

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

library(spm)

data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
longlat <- petrel[, c(1, 2)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)
y <- log(gravel[, 7] +1)

svmkrigepred1 <- svmkrigepred(formula.svm = model, longlat = longlat, trainxy =  gravel,
predx = petrel.grid, y = y, longlatpredx = petrel.grid[, c(1:2)],
transformation = "none", formula.krige = res1 ~ 1, vgm.args = "Sph", nmaxkrige = 12)

names(svmkrigepred1)

# Back transform 'svmkrigepred$predictions' to generate the final predictions
svmkrige.predictions <- exp(svmkrigepred1$predictions) - 1
range(svmkrige.predictions)

Generate spatial predictions using support vector machine ('svm')

Description

This function is for generating spatial predictions using 'svm' method in 'e1071' package.

Usage

svmpred(
  formula = NULL,
  trainxy,
  longlatpredx,
  predx,
  scale = TRUE,
  type = NULL,
  kernel = "radial",
  degree = 3,
  gamma = if (is.vector(trainxy)) 1 else 1/ncol(trainxy),
  coef0 = 0,
  cost = 1,
  nu = 0.5,
  tolerance = 0.001,
  epsilon = 0.1,
  ...
)

Arguments

Value

A dataframe of longitude, latitude and predictions.

Author(s)

Jin Li

References

David Meyer, Evgenia Dimitriadou, Kurt Hornik, Andreas Weingessel and Friedrich Leisch (2020). e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. R package version 1.7-4. https://CRAN.R-project.org/package=e1071.

Examples

library(spm)
data(petrel)
data(petrel.grid)

gravel <- petrel[, c(1, 2, 6:9, 5)]
model <- log(gravel + 1) ~  lat +  bathy + I(long^3) + I(lat^2) + I(lat^3)

svmpred1 <- svmpred(formula = model, trainxy = gravel,
longlatpredx = petrel.grid[, c(1:2)], predx = petrel.grid)

names(svmpred1)

# Back transform 'svmpred1$pred.svm1' to generate the final predictions
svm.predictions <- exp(svmpred1$pred.svm1) - 1
range(svm.predictions)

Cross validation, n-fold and leave-one-out for thin plate splines ('TPS')

Description

This function is a cross validation function for 'Tps' method in 'fields' package.

Usage

tpscv(
  trainx,
  trainy,
  m = NULL,
  p = NULL,
  theta = 3,
  lambda = NULL,
  lon.lat = TRUE,
  validation = "CV",
  cv.fold = 10,
  predacc = "VEcv",
  ...
)

Arguments

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Note

This function is largely based on 'krigecv' in this package and 'Tps' and 'fastTpsMLE' in 'fields' package.

Author(s)

Jin Li

References

Douglas Nychka and Reinhard Furrer and John Paige and Stephan Sain and Florian Gerber and Matthew Iverson, 2020. fields: Tools for Spatial Data, R package version 10.3 https://CRAN.R-project.org/package=fields.

Examples

library(fields)
library(spm)
data(petrel)

tpscv1 <- tpscv(petrel[, c(1,2)], petrel[, 5], cv.fold = 5, predacc = "VEcv")
tpscv1

tpscv1 <- tpscv(petrel[, c(1,2)], petrel[, 5], lambda = 0.9, cv.fold = 5, predacc = "VEcv")
tpscv1

tpscv1 <- tpscv(petrel[, c(1,2)], petrel[, 5], validation = "LOO", predacc = "VEcv")
tpscv1

set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
tpscv1 <- tpscv(petrel[, c(1,2)], petrel[, 5], cv.fold = 10,  lambda = 0.13, predacc = "VEcv")
VEcv [i] <- tpscv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for TPS", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

n <- 20 # number of iterations,60 to 100 is recommended.
# set.seed(1234)
VEcv <- NULL
for (i in 1:n) {
set.seed(1234 + i) # set random seed for each iteration. You can remove
# this line and use above set.seed(1234) and see what you can get.
tpscv1 <- tpscv(petrel[, c(1,2)], petrel[, 5], predacc = "VEcv")
VEcv [i] <- tpscv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for TPS", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)