Date:	2023-01-27
Type:	Package
Title:	Easy Model-Builder Functions for 'OpenMx'
Imports:	methods
Description:	Utilities for building certain kinds of common matrices and models in the extended structural equation modeling package, 'OpenMx'.
Depends:	R (≥ 3.0.0), OpenMx
Suggests:	rpf (≥ 0.45)
License:	GPL-2 | GPL-3 [expanded from: GPL]
Version:	0.3-2
NeedsCompilation:	no
URL:	https://bitbucket.org/mhunter/easymx
BugReports:	https://bitbucket.org/mhunter/easymx/issues
Packaged:	2023-01-30 21:31:39 UTC; mhunter
Author:	Michael D. Hunter [aut, cre], Joshua N. Pritikin [ctb]
Maintainer:	Michael D. Hunter <mhunter.ou@gmail.com>
Repository:	CRAN
Date/Publication:	2023-01-30 22:00:07 UTC

EasyMx: Easy modeling in OpenMx

Description

EasyMx is a package for extended structural equation modeling. It is built as a higher-level frontend on top of OpenMx. It is intended as an Easy introduction to OpenMx: Easy Mx. Try the example below.

Details

All of the functions in the EasyMx package create OpenMx objects. These are most often MxMatrix, MxAlgebra, or MxModel objects. The primary difference between EasyMx and OpenMx is design philosophy. OpenMx has its foundation in WYSIWID: What you say is what it does. This requires the user to be very explicit. The EasyMx package is not as strong or flexible as OpenMx, but it places less burden on the user. Many decisions are made automatically for the user. Some of them are modifiable within EasyMx; for others the user is encouraged to use OpenMx, where nearly everything is modifiable.

The package is broadly divided into two styles of functions: matrix builders and model builders.

The matrix builder functions are utilities for building common structural equation model matrices. These include emxLoadings for factor loadings, emxResiduals for residual variances, emxCovariances for latent or manifest covariances, emxMeans for means and intercepts matrices, and emxThresholds for thresholds matrices when ordinal data are involved.

The model builder functions are higher-level utilities for building common kinds of structural equation models. The model builders often call several matrix builders. The model builders include emxFactorModel for (multiple) factor models, emxGrowthModel for latent growth curve models, and emxRegressionModel for full-information likelihood estimation of regression for observed variables.

There are also a few model builder functions for non-standard structural equation models. In particular, the emxVARModel function creates vector autoregressive models, and the emxStateSpaceMixtureModel function creates state space mixture models.

A third category of functions encompasses special functions for behavior genetics modeling. Some of these functions are matrix builders, and others are model builders. The lowest-level functions for behavior genetics are emxCholeskyVariance, emxGeneticFactorVariance, emxRelatednessMatrix, and emxKroneckerVariance.

A higher-level set of behavior genetics matrix builders create all the matrices and algebraic statements needed for e.g. the A component of an ACE model. These functions are emxCholeskyComponent and emxGeneticFactorComponent.

The highest-level of behavior genetics functions builds some basic twin models. The primary function for this is emxTwinModel.

Finally, a mixture model helper is provided: emxMixtureModel.

Examples

# Make and run a one factor model
## Not run: 
require(EasyMx)

data(demoOneFactor)
fmod <- list(G=names(demoOneFactor))
fit1 <- emxFactorModel(fmod, demoOneFactor, run=TRUE)
summary(fit1)

## End(Not run)

Creates component for a Biometric Cholesky Model

Description

This function creates all the objects needed for a Cholesky component.

Usage

emxCholeskyComponent(x, xname, xvalues, xfree,
                     h=2, hname=paste0('H', xname), hvalues, hlabels)

Arguments

Details

This function is a combination of emxCholeskyVariance, emxRelatednessMatrix, and emxKroneckerVariance.

Value

A list with elements (1) the lower triangular matrix for the Cholesky, (2) the full positive definite variance matrix, (3) the relatedness matrix, and (4) the Kronecker product of the variance matrix and the relatedness matrix.

See Also

emxGeneticFactorComponent

Examples

   
# Create an ACE model in 22 lines
require(EasyMx)
require(OpenMx)
data(twinData)
twinVar = names(twinData)
selVars <- c('ht1', 'bmi1','ht2','bmi2')
mzdzData <- subset(twinData, zyg %in% c(1, 3), c(selVars, 'zyg'))
mzdzData$RCoef <- c(1, NA, .5)[mzdzData$zyg]
nVar = length(selVars)/2
x <- paste0('x', 1:nVar)

acomp <- emxCholeskyComponent(x, 'A', hvalues=c(1, .5, 1), hlabels=c(NA, 'data.RCoef', NA))
ccomp <- emxCholeskyComponent(x, 'C', hvalues=c(1, 1, 1))
ecomp <- emxCholeskyComponent(x, 'E', hvalues=c(1, 0, 1))
totalVar <- mxAlgebra(AKron + CKron + EKron, 'V', dimnames=list(selVars, selVars))
totalMean <- emxMeans(selVars, type='twin')
expect <- mxExpectationNormal(totalVar$name, totalMean$name)
fitfun <- mxFitFunctionML()

comlist <- c(acomp, ccomp, ecomp, list(totalVar, totalMean, expect, fitfun))

model <- mxModel('model', comlist, mxData(mzdzData, 'raw'))
## Not run: 
run2 <- mxRun(model)

## End(Not run)

Create a variance matrix in Cholesky form

Description

This function creates a Cholesky variance matrix and associated MxMatrix and MxAlgebra objects.

Usage

emxCholeskyVariance(x, name, values=.8, free=TRUE)

Arguments

Details

This is a helper function for creating a matrix that is symmetric and positive definite. Full covariance matrices are the most common case of these. In a behavior genetics modeling context, Cholesky components can be created for Additive genetics, Common environments, and unique Environments. These are unrestrictive models of the covariances of multiple phenotypes.

Value

A list with two components. The first component is the lower triangular MxMatrix. The second component is an MxAlgebra, the result of which is the positive definite variance matrix.

See Also

emxGeneticFactorVariance

Examples

   
# Create a Cholesky variance matrix called 'A'
require(EasyMx)
nVar <- 3
x <- paste0('x', 1:nVar)
amat <- emxCholeskyVariance(x, 'A')

Creates component for a Biometric Common Pathway Model

Description

UNDER ACTIVE DEVELOPMENT. DO NOT TRUST. This function creates all the objects needed for a Common Pathway component.

Usage

emxCommonPathwayComponent(x, xname, xvalues=.8, xfree=TRUE, xlbound=NA, xubound=NA,
                          h=2, hname=paste0('H', xname), hvalues, hlabels)

Arguments

Details

This function is a combination of emxCholeskyVariance, emxRelatednessMatrix, and emxKroneckerVariance.

Value

A list with elements (1) the lower triangular matrix for the Cholesky, (2) the full positive definite variance matrix, (3) the relatedness matrix, and (4) the Kronecker product of the variance matrix and the relatedness matrix.

See Also

emxGeneticFactorComponent

Examples

   
# Create an ACE model in 22 lines
require(EasyMx)
require(OpenMx)
data(twinData)
twinVar = names(twinData)
selVars <- c('ht1', 'bmi1','ht2','bmi2')
mzdzData <- subset(twinData, zyg %in% c(1, 3), c(selVars, 'zyg'))
mzdzData$RCoef <- c(1, NA, .5)[mzdzData$zyg]
nVar = length(selVars)/2
x <- paste0('x', 1:nVar)

acomp <- emxCholeskyComponent(x, 'A', hvalues=c(1, .5, 1), hlabels=c(NA, 'data.RCoef', NA))
ccomp <- emxCholeskyComponent(x, 'C', hvalues=c(1, 1, 1))
ecomp <- emxCholeskyComponent(x, 'E', hvalues=c(1, 0, 1))
totalVar <- mxAlgebra(AKron + CKron + EKron, 'V', dimnames=list(selVars, selVars))
totalMean <- emxMeans(selVars, type='twin')
expect <- mxExpectationNormal(totalVar$name, totalMean$name)
fitfun <- mxFitFunctionML()

comlist <- c(acomp, ccomp, ecomp, list(totalVar, totalMean, expect, fitfun))

model <- mxModel('model', comlist, mxData(mzdzData, 'raw'))
## Not run: 
run2 <- mxRun(model)

## End(Not run)

Create a set of covariances

Description

This function creates a covariance matrix as an MxMatrix or MxPath object.

Usage

emxCovariances(x, values, free, path=FALSE, type, name='Variances')

Arguments

Details

Possible values for the type argument are 'independent', 'full', and 'corr'. When type='independent', the remaining arguments are passes to emxResiduals. The values and free arguments are only used when the type argument is 'independent'. For all other cases, they are ignored.

When type='full', a full covariance matrix is created. That is, a symmetric matrix is created with all unique elements freely estimated. The starting values for the variances are all 1; for the covariances, all 0.5.

When type='corr', a full correlation matrix is created. That is, a symmetric matrix is created with all unique elements not on the diagonal freely estimated. The starting values for the correlations are all 0.5. The variances are fixed at 1.

Value

Depending on the value of the path argument, either an MxMatrix or and MxPath object that can be inspected, modified, and/or included in MxModel objects.

See Also

emxFactorModel, emxGrowthModel

Examples

   
# Create a covariance matrix
require(EasyMx)
manVars <- paste0('x', 1:6)
latVars <- paste0('F', 1:2)
emxCovariances(manVars, type='full')
emxCovariances(latVars, type='corr', path=TRUE)

Create a factor model

Description

This function creates a factor model as an MxModel object.

Usage

emxFactorModel(model, data, name, run=FALSE, identification, use, ordinal,
    ..., parameterization=c("lisrel", "ifa"), weight = as.character(NA))
emxModelFactor(model, data, name, run=FALSE, identification, use, ordinal,
    ..., parameterization=c("lisrel", "ifa"), weight = as.character(NA))

Arguments

Details

The model argument must be a named list. The names of the list give the names of the latent variables. Each list element gives the names of the variables that load onto that latent variable. This may sound complicated, but the example below makes this more clear. It is intended to be visually intuitive.

Value

An MxModel.

See Also

emxLoadings

Examples

   
# Example
require(EasyMx)
data(myFADataRaw)
xmap <- list(F1=paste0('x', 1:6), F2=paste0('y', 1:3), F3=paste0('z', 1:3))
## Not run: 
mod <- emxFactorModel(xmap, data=myFADataRaw, run=TRUE)

## End(Not run)

Creates component for a Genetic Factor Model

Description

This function creates all the objects needed for Genetic Factor component

Usage

emxGeneticFactorComponent(x, xname, xvalues=.8, xfree=TRUE, xlbound=NA, xubound=NA,
                          h=2, hname=paste0('H', xname), hvalues, hlabels)

Arguments

Details

This function is a combination of emxGeneticFactorVariance, emxRelatednessMatrix, and emxKroneckerVariance.

Value

A list with elements (1) the genetic factor loadings matrix, (2) the full positive definite variance matrix, (3) the relatedness matrix, and (4) the Kronecker product of the variance matrix and the relatedness matrix.

See Also

emxCholeskyComponent

Examples

   
# Create genetic factor A component for DZ twins
require(EasyMx)
xvars <- paste0('x', 1:4)
acomp <- emxGeneticFactorComponent(xvars, 'A', hvalues=c(1, .5, 1))

Creates a variance matrix accoring to the Genetic Factor Model

Description

This function creates a variance matrix according to the genetic factor model

Usage

emxGeneticFactorVariance(x, name, values=.8, free=TRUE, lbound=NA, ubound=NA)

Arguments

Value

A list with two components. The first component is the factor loadings matrix. The second component is an MxAlgebra, the result of which is the variance matrix implied by the factor loadings.

See Also

emxCholeskyVariance

Examples

   
# Create a genetic factor variance matrix
require(EasyMx)
xvars <- paste0('x', 1:2)
emxGeneticFactorVariance(xvars, 'D')

Create a latent growth curve model

Description

This function creates a latent growth curve model as an MxModel object.

Usage

emxGrowthModel(model, data, name, run=FALSE, identification, use, ordinal, times)
emxModelGrowth(model, data, name, run=FALSE, identification, use, ordinal, times)

Arguments

Details

The model argument can be either a character or a number that tells the kind of growth curve to make. If it is a character it currently must be one of "Intercept", "Linear", "Quadratic", "Cubic", "Quartic", or "Quintic", and it produces a polynomial growth curve of the corresponding type. If it is a number, the function produces a polynomial growth curve of the corresponding order. Zero is an intercept only, one is linear, two is quadratic; and so on.

When missing, the times are assumed to start at zero and increment by one until the number of variables is completed.

Value

An MxModel

See Also

emxFactorModel, emxGrowthModel

Examples

   
# Example
require(EasyMx)
data(myLongitudinalData)
## Not run: 
mod <- emxGrowthModel('Linear', data=myLongitudinalData, use=names(myLongitudinalData), run=TRUE)

## End(Not run)

Creates component for a Biometric Independent Pathway Model

Description

UNDER ACTIVE DEVELOPMENT. DO NOT TRUST. This function creates all the objects needed for an Independent Pathway component.

Usage

emxIndependentPathwayComponent(x, xname, xvalues=.8, xfree=TRUE, xlbound=NA, xubound=NA,
                               h=2, hname=paste0('H', xname), hvalues, hlabels)

Arguments

Details

This function is a combination of emxCholeskyVariance, emxRelatednessMatrix, and emxKroneckerVariance.

Value

A list with elements (1) the lower triangular matrix for the Cholesky, (2) the full positive definite variance matrix, (3) the relatedness matrix, and (4) the Kronecker product of the variance matrix and the relatedness matrix.

See Also

emxGeneticFactorComponent

Examples

   
# Create an ACE model in 22 lines
require(EasyMx)
require(OpenMx)
data(twinData)
twinVar = names(twinData)
selVars <- c('ht1', 'bmi1','ht2','bmi2')
mzdzData <- subset(twinData, zyg %in% c(1, 3), c(selVars, 'zyg'))
mzdzData$RCoef <- c(1, NA, .5)[mzdzData$zyg]
nVar = length(selVars)/2
x <- paste0('x', 1:nVar)

acomp <- emxCholeskyComponent(x, 'A', hvalues=c(1, .5, 1), hlabels=c(NA, 'data.RCoef', NA))
ccomp <- emxCholeskyComponent(x, 'C', hvalues=c(1, 1, 1))
ecomp <- emxCholeskyComponent(x, 'E', hvalues=c(1, 0, 1))
totalVar <- mxAlgebra(AKron + CKron + EKron, 'V', dimnames=list(selVars, selVars))
totalMean <- emxMeans(selVars, type='twin')
expect <- mxExpectationNormal(totalVar$name, totalMean$name)
fitfun <- mxFitFunctionML()

comlist <- c(acomp, ccomp, ecomp, list(totalVar, totalMean, expect, fitfun))

model <- mxModel('model', comlist, mxData(mzdzData, 'raw'))
## Not run: 
run2 <- mxRun(model)

## End(Not run)

Creates a large Variance matrix by Kroneckering two smaller matrices

Description

This function creates the wide format variance matrix when combined with a relatedness matrix

Usage

emxKroneckerVariance(h, v, name)

Arguments

Details

In many behavior genetic models, a relationship matrix is combined with a base variance matrix. The combination is done with a Kronecker product so that the variance exists (possibly weighted by zero or another number) for each member of the relationship.

Value

MxAlgebra

See Also

emxRelatednessMatrix

Examples

   
# Create a loadings matrix
require(EasyMx)
x <- letters[23:26]
amat <- emxCholeskyVariance(x, 'A')
ahmat <- emxRelatednessMatrix(2, c(1, .5, 1), name='AH')
ab <- emxKroneckerVariance(ahmat, amat[[2]], 'AB')

Create a factor loadings matrix

Description

This function creates a factor loadings matrix as an MxMatrix or MxPath object.

Usage

emxLoadings(x, values=.8, free=TRUE, path=FALSE)

Arguments

Details

The x argument must be a named list. The names of the list give the names of the latent variables. Each list element gives the names of the variables that load onto that latent variable. This may sound complicated, but the example below makes this more clear. It is intended to be visually intuitive.

Value

Depending on the value of the path argument, either an MxMatrix or and MxPath object that can be inspected, modified, and/or included in MxModel objects.

See Also

emxFactorModel

Examples

   
# Create a loadings matrix
require(EasyMx)
xmap <- list(F1=paste0('x', 1:6), F2=paste0('y', 1:3), F3=paste0('z', 1:3))
emxLoadings(xmap)
emxLoadings(xmap, path=TRUE)

Create a set of means

Description

This function creates a means matrix as an MxMatrix or MxPath object.

Usage

emxMeans(x, values=0, free=TRUE, path=FALSE, type='saturated', name, column=TRUE, labels)

Arguments

Details

Possible values for the type argument are 'saturated', 'equal', 'twin', 'special'.

Value

Depending on the value of the path argument, either an MxMatrix or and MxPath object that can be inspected, modified, and/or included in MxModel objects.

See Also

emxFactorModel, emxGrowthModel

Examples

   
# Create a covariance matrix
require(EasyMx)
manVars <- paste0('x', 1:6)
emxMeans(manVars, type='saturated')

Create a mixture model

Description

This function creates a mxiture model as an MxModel object.

Usage

emxMixtureModel(model, data, run=FALSE, p=NA, ...)
emxModelMixture(model, data, run=FALSE, p=NA, ...)

Arguments

Details

The model argument is list of MxModel objects. These are the classes over which the mixture model operates.

The p argument is optional. If not specified, the function will create and properly scale the mixing proportions for you. If specified, the Mx Object that gives the mixing proportions should be a column vector (one-column matrix).

Value

An MxModel.

See Also

emxLoadings

Examples

   
# Factor Mixture Example
require(EasyMx)
data(myFADataRaw)
xmap1 <- list(F1=paste0('x', 1:6), F2=paste0('y', 1:3), F3=paste0('z', 1:3))
mod1 <- emxFactorModel(xmap1, data=myFADataRaw, name='m1')

xmap2 <- list(F1=c(paste0('x', 1:6), paste0('y', 1:3), paste0('z', 1:3)))
mod2 <- emxFactorModel(xmap2, data=myFADataRaw, name='m2')

mod <- emxMixtureModel(list(mod1, mod2), data=myFADataRaw)
# To estimate parameters either
#  1. mod <- mxRun(mod)   or
#  2. include run=TRUE in the arguments above
summary(mod)
coef(mod)

# Latent Profile Example
require(EasyMx)

m1 <- omxSaturatedModel(demoOneFactor)[[1]]
m1 <- mxRename(m1, 'profile1')

m2 <- omxSaturatedModel(demoOneFactor)[[1]]
m2 <- mxRename(m2, 'profile2')

mod <- emxMixtureModel(list(m1, m2), data=demoOneFactor)
# To estimate parameters either
#  1. mod <- mxRun(mod)   or
#  2. include run=TRUE in the arguments above
summary(mod)
coef(mod)

mxGetExpected(mod$profile1, 'covariance')
mxGetExpected(mod$profile1, 'means')
mxGetExpected(mod$profile2, 'covariance')
mxGetExpected(mod$profile2, 'means')

Create a regression model

Description

This function creates a regression model as an MxModel object.

Usage

emxRegressionModel(model, data, type='Steven', run, ...)
emxModelRegression(model, data, type='Steven', run, ...)

Arguments

Details

The model argument is a formula identical to what is used in lm.

The type argument switches the kind of regression model that is specified. When there are no missing data, the two versions will estimate the same regression parameters but type='Steven' will estimate addition parameters that are not estimated by type='Joshua'. The type='Steven' model is due to Steven Boker and many others. It estimates more parameters than a typical regression analysis and has a different set of assumptions. More exactly, type='Steven' models the outcome and all of the predictors as a multivariate Normal distribution. By contrast, type='Joshua' is due to Joshua Pritikin and exactly replicates the typical regression model with its usual assumptions. In particular, type='Joshua' models the regression residual as a univariate Normal distribution. Predictors are assumed to have no measurement error (see Westfall & Yarkoni, 2016).

The benefit of type='Steven' is that it handles missing data with full-information maximum likelihood (FIML; Enders & Bandalos, 2001), at the cost of using a different model with different assumptions from ordinary least squares regression. The benefit of type='Joshua' is that it exactly replicates regression as a maximum likelhood model, at the cost of having the same weakness in terms of missing data as OLS regression.

Value

An MxModel.

References

Enders, C. K. & Bandalos, D. L. (2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. <i>Structural Equation Modeling, 8</i>(3), 430-457.

Westfall, J. & Yarkoni, T. (2016). Statistically controlling for confounding constructs is harder than you think. <i>PLoS ONE, 11</i>(3). doi:10.1371/journal.pone.0152719

See Also

lm

Examples

   
# Example
require(EasyMx)
data(myRegDataRaw)
myrdr <- myRegDataRaw
myrdr[1, 4] <- NA

## Not run: 
run <- emxRegressionModel(y~1+x*z, data=myrdr, run=TRUE)
summary(run)

## End(Not run)

summary(lm(y~1+x*z, data=myrdr))

Create a relatedness matrix

Description

This function creates a relatedness matrix as an MxMatrix, often used in behavior genetics modeling.

Usage

emxRelatednessMatrix(nvar, values, labels, name='h')

Arguments

Details

The labels argument can be used to create a "definition variable" which populates the value from one of the data columns for each row in the data. In this context, if the genetic relatedness coefficient between a pair of individuals is given by a column in the data then that information can be used to create in the relatedness matrix. Alternatively, multiple groups can be created

Value

An MxMatrix object.

See Also

emxGeneticFactorVariance

Examples

   
# Create a Cholesky variance matrix called 'A'
require(EasyMx)
ahmat <- emxRelatednessMatrix(2, c(1, .5, 1), labels=c(NA, 'data.RCoef', NA), name='AH')
# data.RCoef creates a definition variable and ignores the .5 value.

Create a residual variances matrix

Description

This function creates a factor loadings matrix as an MxMatrix or MxPath object.

Usage

emxResiduals(x, values=.2, free=TRUE, lbound=NA, ubound=NA, path=FALSE, type='unique')

Arguments

Details

Possible values for the type argument are 'unique' and 'identical'. When type='unique', each residual variances is a unique free parameter. When type='identical', all of the residual variances are given by a single free parameter. In this case, all the residual variances are constrained to be equal. However, no linear or non-liniear contraint function is used. Rather, a single parameter occurs in multiple locations by using the same label.

Value

Depending on the value of the path argument, either an MxMatrix or and MxPath object that can be inspected, modified, and/or included in MxModel objects.

See Also

emxFactorModel, emxGrowthModel

Examples

   
# Create a residual variance matrix
require(EasyMx)
manVars <- paste0('x', 1:6)
emxResiduals(manVars, lbound=1e-6)
emxResiduals(manVars, type='identical')
emxResiduals(manVars, path=TRUE)

Classify time series in a state space mixture model

Description

This function classifies time series (usually people) in a state space mixture model.

Usage

emxStateSpaceMixtureClassify(model)

Arguments

Details

This is a helper function for state space mixture modeling. The function will almost exclusively be used in conjunction with emxStateSpaceMixtureModel. The present function takes a state space mixture model as input, and returns detailed information about the most likely class for each unique ID.

Value

A named list with elements

See Also

emxStateSpaceMixtureModel , emxMixtureModel

Examples

   
# Example
require(EasyMx)
data(myFADataRaw)

ds0 <- myFADataRaw[,1:3]

# Make a VAR Model
vm <- emxVARModel(data=ds0, use=names(ds0), name='varmodel')

# Re-label parameters to have different AR parameters
# for class 1 and class 2
vm1 <- OpenMx::omxSetParameters(vm, labels=vm$Dynamics$labels,
    newlabels=paste0(vm$Dynamics$labels, '_k1'), name='klass1')
vm2 <- OpenMx::omxSetParameters(vm, labels=vm$Dynamics$labels,
    newlabels=paste0(vm$Dynamics$labels, '_k2'), name='klass2')

# Pretend you have a data set of 50 people
# each measured 10 times on 3 variables
ds1 <- myFADataRaw[, 1:3]
ds1$id <- rep(1:50, each=nrow(myFADataRaw)/50)

## Not run: 
# Make the state space mixture model
ssmm <- emxStateSpaceMixtureModel(model=list(vm1, vm2),
    data=ds1, ID='id')

# Fit model
ssmmr <- mxRun(ssmm)

# Extract estimated classes and diagnostics
eclasses <- emxStateSpaceMixtureClassify(ssmmr)

## End(Not run)

Create a state space mixture model

Description

This function creates a state space mixture model as an MxModel object.

Usage

emxStateSpaceMixtureModel(model, data, name, run=FALSE, use, ID, time, ...)
emxModelStateSpaceMixture(model, data, name, run=FALSE, use, ID, time, ...)

Arguments

Details

The idea of state space mixture modeling is to model multiple, multivariate time series while allowing for qualitative differences between the time series. Suppose you have a multivariate time series for several people. You think some people should have the same time series model, but not everyone. You think there should be a small number of homogeneous sets of people that follow the same time series model, but you do not know which people or the exact parameter values of the candidate time series models. This function presents one solution to this problem.

State space mixture modeling begins with a set of candidate state space models, and uses these state space models as the mixture classes. The goal is to simultaneously estimate the free parameters of the state space models, and estimate which multivariate time series (e.g., perseon) belongs to which mixture class.

The component state space models may share some free parameters or none. Note that free parameters with the same name are constrained to be equal across all models. Conversely, unnamed free parameters are given unique names and are allowed to differ for each person-mixture combination, which creates a very large number of free parameters. We strongly encourage you to name all of your free parameters in the model list to avoid melting your computer's CPU.

The model argument currently must be a list. The elements of the list should be MxModel objects. Each list element forms a mixture class in the final model.

This function creates a multigroup mixture model where the mixture classes are the elements of the model list argument. Each unique ID forms a group.

The data argument can be a list of data.frame objects with one element for each ID, or a single data.frame with an ID variable that separates groups.

Value

An MxModel.

See Also

emxStateSpaceMixtureClassify , emxMixtureModel

Examples

   
# Example
require(EasyMx)
data(myFADataRaw)

ds0 <- myFADataRaw[,1:3]

# Make a VAR Model
vm <- emxVARModel(data=ds0, use=names(ds0), name='varmodel')

# Re-label parameters to have different AR parameters
# for class 1 and class 2
vm1 <- OpenMx::omxSetParameters(vm, labels=vm$Dynamics$labels,
    newlabels=paste0(vm$Dynamics$labels, '_k1'), name='klass1')
vm2 <- OpenMx::omxSetParameters(vm, labels=vm$Dynamics$labels,
    newlabels=paste0(vm$Dynamics$labels, '_k2'), name='klass2')

# Pretend you have a data set of 50 people
# each measured 10 times on 3 variables
ds1 <- myFADataRaw[, 1:3]
ds1$id <- rep(1:50, each=nrow(myFADataRaw)/50)

## Not run: 
# Make and fit the state space mixture model
ssmm <- emxStateSpaceMixtureModel(model=list(vm1, vm2),
    data=ds1, ID='id', run=TRUE)

## End(Not run)

Create a set of thresholds for ordinal data

Description

This function creates a threshold matrix as an MxMatrix object.

Usage

emxThresholds(data, ordinalCols, deviation=TRUE)

Arguments

Value

An MxMatrix giving the thresholds.

See Also

emxFactorModel, emxGrowthModel

Examples

   
# Example
require(EasyMx)
data(jointdata)
jointdata[, c(2, 4, 5)] <- mxFactor(jointdata[,c(2, 4, 5)],
	levels=sapply(jointdata[,c(2, 4, 5)], function(x){sort(unique(x))}))
emxThresholds(jointdata, c(FALSE, TRUE, FALSE, TRUE, TRUE))

Creates behavior genetics Twin Model

Description

This function creates an MxModel and associated objects for a basic Twin model.

Usage

emxTwinModel(model, relatedness, data, run=FALSE, use, name='model', components='ACE')
emxModelTwin(model, relatedness, data, run=FALSE, use, name='model', components='ACE')

Arguments

Details

Because the model argument is ignored and the relatedness argument has limited use, this function only constructs a very basic and rigid Twin model. It creates a Cholesky model with A, C, and E components or a Cholesky model with A, D, and E components. The means are constrained equal across twins.

Value

MxModel.

See Also

emxFactorModel

Examples

   
# Create an ACE model in 10 lines
# 8 of those are data handling.
# 2 are the actual model.
require(EasyMx)
require(OpenMx)
data(twinData)
twinVar = names(twinData)
selVars <- c('ht1', 'bmi1','ht2','bmi2')
mzdzData <- subset(twinData, zyg %in% c(1, 3), c(selVars, 'zyg'))
mzdzData$RCoef <- c(1, NA, .5)[mzdzData$zyg]

## Not run: 
run3 <- emxTwinModel(model='Cholesky', relatedness='RCoef',
	data=mzdzData, use=selVars, run=TRUE, name='TwCh')

## End(Not run)

Create a vector autoregressive (VAR) model

Description

This function creates a vector autoregressive (VAR) model as an MxModel object.

Usage

emxVARModel(model, data, name, run=FALSE, use, ID)
emxModelVAR(model, data, name, run=FALSE, use, ID)

Arguments

Details

The purpose of this function is to quickly specify a vector autoregressive model. It is currently in the early stages of development and might change considerable with regard to the model argument and the ID argument.

Value

An MxModel.

See Also

emxStateSpaceMixtureModel , emxFactorModel

Examples

   
# Example
require(EasyMx)
data(myFADataRaw)
ds0 <- myFADataRaw[,1:3]

# Make a VAR Model
vm <- emxVARModel(data=ds0, use=names(ds0), name='varmodel')

Creates Variance Components Model

Description

This function creates a variance components model as an MxModel object.

Usage

emxVarianceComponents(model, data, run)

Arguments

Details

This function does not really do anything currently. Do not use it.

Value

MxModel.

See Also

emxFactorModel

Examples

   
# Create a loadings matrix
require(EasyMx)