The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
oneStep, triangle, and huber) is
now accomplished by OSCARS::oscars. OSCARS is a true global
maximizer and hence cannot absolutely guarantee that the global maximum
has been found. However, OSCARS now starts at parameters reasonably
close to the likelihood’s maximum and performs a maximum of 10,000
iterations (the default, see
options("Rdistance_oscarEvals")). While slower than other
methods, this method usually finds a higher maximum of the likelihood
than other methods. Due to the significant slow down using
OSCARS, estimated run times are reported on the command line.findSpacing() computes transect spacing that yields a
target survey length. Routine makeLines() places transects
with a random start. calcLineLength() converts a target
number of detected groups into a target transect length. The main
function called by users, drawTransects(), is a wrapper
that calls findSpacing() followed by
makeLines(). Both parallel (“rectangular”, or “back and
forth”) and “zigzag” transects are implemented. Transects are applied
over one or several polygons, with an option to generate random
replicates (parameter R). Transects can be returned as one
continuous route (combine = TRUE) or as
individual legs (combine = FALSE), and
targetLength can refer to the total (target =
"total") or on-effort length (target =
"onEffort"). Computation of the spacing is an optimization
problem performed by OSCARS::oscars. Some notes:
combine controls
clipping and shape of the returned object. combine = TRUE
returns the full route, unclipped, including on-transect and
off-transect transit segments. combine = FALSE returns one
row per “leg”, clipped to the polygon. The returned lengths sum to
on-effort length. Legs that are broken by a concavity is divided into
pieces and is returned as a MULTILINESTRING geometry.baseline is supplied, the polygon’s general centerline is
approximated by the midpoints of polygon cords. This initial baseline is
then straightened by regressing vertical coordinates onto horizontal
coordinates and translating the estimated line to pass through the
polygon’s centroid.convexPartition splits a strongly concave polygon into a
number of (hopefully) more-convex pieces using Approximate Convex
Decomposition (Lien & Amato 2006). Splitting improves the coverage
of zigzag transects on bent, arc-, or L-shaped polygons. Breaking
polygons into smaller more-convex pieces is not necessary for
rectangular transects. Parameter nPieces is either an
integer, in which case exactly that many pieces are returned, or
"optimum" (the default), in which case the count is chosen
automatically. method = "fast" (the default) cuts greedily
at the most concave vertex. method = "optimum" searches the
cut vertices with OSCARS::oscars to maximize the minimum
solidity of the pieces.exampleSurveyPoly: two real-world concave survey strata
in Alaska (in Cook Inlet) used to demonstrate the survey-design
functions.pronghornDf: Data from an aerial line-transect survey
for pronghorn (Antilocapra americana) collected by the Wyoming
Game and Fish Department in southeast Wyoming, 2012-2019.pronghornAreas: The study-area (herd-unit) polygons
associated with the pronghorn line-transect data.sf,
grDevices, and OSCARS to Imports, required by
the new survey-design functions.abundEstim() now
accepts a previously fitted abundance object in addition to a distance
function. This lets users add bootstrap iterations to an existing fit,
for example by fitting with ci = NULL first and running (or
extending) the bootstrap later.integrateOneStepNumeric
returned areas without units, which was okay for optimization, but not
okay for ESW. integrateOneStepNumeric now returns a vector
with the correct units and is consistent with the other integrate???
functions.triangle
distance function, a mixture of triangle and uniform distributions. See
help(triangle.like) examples.huber
distance function, a mixture of an inverted version of Huber loss and a
uniform distribution. See help(huber.like) examples.oneStep Theta parameter to the minimum and maximum of
distances, thus ensuring the ledge lies within the range of observed
distances.stats::optim for optimization. Prior to this, the only
non-gradient optimizer was HookeJeeves. Default optimizer for
oneStep, triangle, and huber is
now the Nelder-Mead implementation in optim.g.x.scl) were less than 1.0. Versions <=4.1.0 reported
correct ESW, and ESW reported by 4.3.0 was correct when
g.x.scl equaled 1.0 (the default).w.lo changed in
parseModel to suppress warning re x.scl <
w.lo.multidplyr. Default is to run bootstraps in
parallel on n-1 CPU cores, where n is the number of
cores available on the local machine . No progress bar is produced
during parallel processing. Disable parallel processing by setting
parallel = FALSE in call to abundEstim. Run on
a specific number of cores by setting parallel to that
number (e.g., parallel = 3 uses three cores).oneStep.like, a mixture of non-overlapping uniform
densities, as a distance function. Included associated print, plot,
summary, and expansion methods. This required inclusion of a new
non-gradient based optimizer.%#% operator
(e.g., 3 %#% “m”), which makes unit assignment easier than in prior
versions (which used units::set_units). Fixed unit
assignment operators are included for all popular linear and squared
units (e.g., 3 %m%. assigns meters to 3). See
help(unitHelpers).cosine, hermite, simple, and
bspline.options(Rdistance_verbocity = 1) or
higher to see progressively more detailed intermediate output.plot.dfunc.para when w.lo > 0 . Fixed bug
in point transect methods resulting in incorrect likelihood scalings .
Fixed expansions hermite and simple that was
causing non-convergence issuesBug fixes:
predict method when
type = "density" causing NaN estimates on transects with
observations outside the strip.RdistDf when merge parameter
by was named. When by was named and merge was
on different named variables, names and values in by were
reversed prior to the fix due to first nesting then merging. This bug
did not affect merges on same-named variables.Methods and workflow in Rdistance versions >4.0.0 are substantially different from prior versions.
RdistDf constructs the new Rdistance data frames from
separate transect and detection data frames formatted for use in prior
versions. Use this function on old sets of site and detection data
frames to construct the new nested data frames. See examples in
?RdistDf.is.RdistDf checks the validity of the new Rdistance
data frames.summary prints a summary of number of transects, number
of groups seen, number of individuals, etc.print, summary, and plot
methods have been improved.options function. All Rdistance options are prefixed with
‘Rdistance_’ to distinguish them from other options.df |> dfuncEstim(dist~1) |> abundEstim(), which will
estimate a distance function and density in one go.Version 3.1.3 contains three patches. Several updates to documentation and one bug fix precipitated by changes to the ‘units’ package.
Version 3.1.0 primarily addresses GitHub issues.
summary methods for dfunc and
abund objects prior to version v3.1.0. This version implements both
summary and print methods for the main
outputs. print.dfunc and print.abund are
modeled on print methods for lm and glm objects. New methods
summary.dfunc and summary.abund are modeled on
the summary methods for lm and glm and will produce fuller (relative to
print) outputs.abundEstim was called with tibbles.Version 3.0.0 is a substantial change and upgrade.
units package, is internal, and
automatic.singleSided =
TRUE in call to abundEstim.abundEstim, and it is included in bootstrapping.uniform likelihood to logistic.
Uniform is now deprecated.area in abundEstim from 1 to
NULL. NULL now translates to 1 square output unit.abund object.F.gx.estim that occasionally popped up
when sighting function was monotonically decreasing.negexp likelihood parameter to
achieve more valid fits.model.matrixx.scl
and g.x.sxl)ESW for w.lo > 0These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.