R/enumRsimulation.R
In smasongarrison/enumR: Factor Enumeration Functions
Defines functions
enumRsimulation
Documented in
enumRsimulation
#' @title enumR simulation function
#'
#' @description This function simulates factor analyses based on numerous parameters, listed below.
#'
#' @param cor What kind of correlation to find, defaults to covariance matrix, but see fa for the choices
#' @param effectmatrix OPTIONAL matrix of item loadings by factor. If not provided, one is generated from \code{items} and \code{nfactors}.
#' @param items_p_f number of items per factor
#' @param itemsR_p_f number of reverse-scored items per factor
#' @param custom_item If TRUE, overrides \code{items_p_f} and \code{itemsR_p_f}. Primary used for importing from shiny interface.
#' @param items OPTIONAL vector of total number of items per factor
#' @param loading Factor loading magnitude. Default is .5
#' @param loading_norm If TRUE, generate factor loadings that average to \code{loading}
#' @param loading_norm_sd If \code{loading_norm} is TRUE, standard deviation of loadings.
#' @param method factoring method – fm="pa" Principal Axis Factor Analysis, fm = "minres" minimum residual (OLS) factoring fm="ml" Maximum Likelihood FA, fm="pc" Principal Components"
#' @param samplesize Number of observations
#' @param ndatasets Number of datasets generated
#' @param nfactors Number of factors to extract.
#' @param patternmatrix OPTIONAL matrix of factor loadings. If not provided, one is generated from \code{rfactors} and \code{nfactors}.
#' @param rfactors Intercorrelations between factors. Default is 0
#' @param r_norm If TRUE, generate normally distributed factor loadings with mean \code{loading} and sd \code{r_norm_sd}. Does not check if correlation matrix is positive definite.
#' @param r_norm_sd If \code{r_norm} is TRUE, standard deviation for normally distributed intercorrelations.
#' @param itemsR Number of items per factor that are reverse scored
#' @param rotation what rotation to use c("none", "varimax", "oblimin","promax")
#' @param seed numeric value for setting a seed. Allows results to be reproduced.
#' @param true_model If TRUE, Estimates model using \code{nfactors} only. Is used in \code{enumRsimulation} function. Default is FALSE.
#' @param use If doing covariances or Pearson R, should we use "pairwise" or "complete cases"
#' @param f1_items Total number of items on factor 1
#' @param f2_items Total number of items on factor 2
#' @param f3_items Total number of items on factor 3
#' @param f4_items Total number of items on factor 4
#' @param f5_items Total number of items on factor 5
#' @param f6_items Total number of items on factor 6
#' @param f7_items Total number of items on factor 7
#' @param f8_items Total number of items on factor 8
#' @param f9_items Total number of items on factor 9
#' @param f10_items Total number of items on factor 10
#' @param f1_itemsR Number of reverse-scored items on factor 1
#' @param f2_itemsR Number of reverse-scored items on factor 2
#' @param f3_itemsR Number of reverse-scored items on factor 3
#' @param f4_itemsR Number of reverse-scored items on factor 4
#' @param f5_itemsR Number of reverse-scored items on factor 5
#' @param f6_itemsR Number of reverse-scored items on factor 6
#' @param f7_itemsR Number of reverse-scored items on factor 7
#' @param f8_itemsR Number of reverse-scored items on factor 8
#' @param f9_itemsR Number of reverse-scored items on factor 9
#' @param f10_itemsR Number of reverse-scored items on factor 10
#' ... parameters to pass to the factor analysis program The most important of these is if using a correlation matrix is covmat= xx
#' @return dataframe
enumRsimulation<-function(seed=12345,
ndatasets=200,
patternmatrix=NULL,
effectmatrix=NULL,
nfactors =5,
loading=.5,
items=NULL,
items_p_f=5,
itemsR_p_f=2,
itemsR=NULL,
loading_norm=FALSE,
loading_norm_sd=.05,
# effects matrix
rfactors=0,
r_norm =FALSE,
r_norm_SD =.015,
samplesize=300,
method="ml",
rotation="oblimin",
custom_item=FALSE,
f1_items=NULL,
f2_items = NULL,
f3_items = NULL,
f4_items = NULL,
f5_items = NULL,
f6_items = NULL,
f7_items = NULL,
f8_items = NULL,
f9_items = NULL,
f10_items = NULL,
f1_itemsR= NULL,
f2_itemsR = NULL,
f3_itemsR = NULL,
f4_itemsR = NULL,
f5_itemsR = NULL,
f6_itemsR = NULL,
f7_itemsR = NULL,
f8_itemsR = NULL,
f9_itemsR = NULL,
f10_itemsR = NULL
){
###prep
set.seed(seed)
require(psych)
require(GPArotation)
require(GAIPE)
simulation<-ndatasets
# cleanup items
if(is.null(items)){
if(custom_item){
items=c(f1_items,
f2_items,
f3_items,
f4_items,
f5_items,
f6_items,
f7_items,
f8_items,
f9_items,
f10_items)
itemsR=c(f1_itemsR,
f2_itemsR,
f3_itemsR,
f4_itemsR,
f5_itemsR,
f6_itemsR,
f7_itemsR,
f8_itemsR,
f9_itemsR,
f10_itemsR)
}else{
items=rep(items_p_f,nfactors)
itemsR=rep(itemsR_p_f,nfactors)
}}
## Results
meganames<-c( "dof","chisq","prob","sqresid","fit", "RMSEA","RMSEA_lower","RMSEA_upper","BIC","SABIC","null.model","null.dof","null.chisq","complex","eChisq","SRMR","eCRMS","eBIC","TLI","fa.fit","fa.fit.off","objective","ENull","eSABIC","null.chisq1" , "TLI.m","NFI.m","CFI.m","RMSEA_lower.m","RMSEA_upper.m","eRMS","cfit.1","cfit.2","cfit.3","cfit.4","cfit.5","cfit.6","cfit.7","cfit.8","cfit.9","cresidual.1","cresidual.2","cresidual.3","cresidual.4","cresidual.5","cresidual.6","cresidual.7","cresidual.8","cresidual.9","map.values","rfactors","samplesize","loadings","dataset","error","factor")
mega<-data.frame(matrix(rep(0,length(meganames)),nrow=1))
names(mega)=meganames
mega_null<-mega
mega_null$error<- 1
percentiles<-data.frame(c(1))
names(percentiles)<-"ds"
if(is.null(patternmatrix)){
fmodel=GenFactorMatrix(nfactors=nfactors, items= items[1:nfactors], itemsR=itemsR[1:nfactors], loading=loading,loading_norm=loading_norm,loading_norm_sd=loading_norm_sd)
}else{fmodel=patternmatrix}
if(is.null(effectmatrix)){
smodel=GenStructure(nfactors=nfactors,rfactors=rfactors,r_norm=r_norm,r_norm_sd=r_norm_sd)
}else{smodel=effectmatrix}
## Gen Data
for(d in 1:simulation) {
data=GenData(fmodel=fmodel,effect=smodel,samplesize=samplesize)
cov<-cov(data)
test<- try(results<- enumR(cov,n=nfactors,samplesize=samplesize,covar=TRUE,truemodel = TRUE,fm = method), silent=TRUE)
if ('try-error' %in% class(test)) {mega[d,]<- mega_null
} else {mega[d,]<- results
mega$factor[d]<-nfactors
mega$error[d]<- 0
}
mega$dataset[d]<-d
print(d)
}
mega$rfactors<-rfactors
mega$samplesize<-samplesize
mega$loadings<-loading
ds_true<-mega[mega$factor==nfactors,]
percentiles$RMSEA_.95<-quantile(ds_true$RMSEA,.95)
percentiles$MAP_.95<-quantile(ds_true$map.values,.95)
#percentiles$chisq_.95<-quantile(ds_true$chisq,.95)
#percentiles$BIC_.95<-quantile(ds_true$BIC,.95)
percentiles$SRMR_.95<-quantile(ds_true$SRMR,.95)
#percentiles$SABIC_.95<-quantile(ds_true$SABIC,.95)
#percentiles$TLI_.95<-quantile(ds_true$TLI,.95)
percentiles$chisq_sig_.95<-quantile(ds_true$prob,.95)
#percentiles$RMSEA_.05<-quantile(ds_true$RMSEA,.05)
#percentiles$MAP_.05<-quantile(ds_true$map.values,.05)
#percentiles$chisq_.05<-quantile(ds_true$chisq,.05)
#percentiles$BIC_.05<-quantile(ds_true$BIC,.05)
#percentiles$SRMR_.05<-quantile(ds_true$SRMR,.05)
#percentiles$SABIC_.05<-quantile(ds_true$SABIC,.05)
percentiles$TLI_.05<-quantile(ds_true$TLI,.05)
percentiles$chisq_sig_.05<-quantile(ds_true$prob,.05)
return(list(percentiles,mega))
}
smasongarrison/enumR documentation built on Feb. 3, 2021, 10:03 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions enumRsimulation
Documented in enumRsimulation
#' @title enumR simulation function
#'
#' @description This function simulates factor analyses based on numerous parameters, listed below.
#'
#' @param cor What kind of correlation to find, defaults to covariance matrix, but see fa for the choices
#' @param effectmatrix OPTIONAL matrix of item loadings by factor. If not provided, one is generated from \code{items} and \code{nfactors}.
#' @param items_p_f number of items per factor
#' @param itemsR_p_f number of reverse-scored items per factor
#' @param custom_item If TRUE, overrides \code{items_p_f} and \code{itemsR_p_f}. Primary used for importing from shiny interface.
#' @param items OPTIONAL vector of total number of items per factor
#' @param loading Factor loading magnitude. Default is .5
#' @param loading_norm If TRUE, generate factor loadings that average to \code{loading}
#' @param loading_norm_sd If \code{loading_norm} is TRUE, standard deviation of loadings.
#' @param method factoring method – fm="pa" Principal Axis Factor Analysis, fm = "minres" minimum residual (OLS) factoring fm="ml" Maximum Likelihood FA, fm="pc" Principal Components"
#' @param samplesize Number of observations
#' @param ndatasets Number of datasets generated
#' @param nfactors Number of factors to extract.
#' @param patternmatrix OPTIONAL matrix of factor loadings. If not provided, one is generated from \code{rfactors} and \code{nfactors}.
#' @param rfactors Intercorrelations between factors. Default is 0
#' @param r_norm If TRUE, generate normally distributed factor loadings with mean \code{loading} and sd \code{r_norm_sd}. Does not check if correlation matrix is positive definite.
#' @param r_norm_sd If \code{r_norm} is TRUE, standard deviation for normally distributed intercorrelations.
#' @param itemsR Number of items per factor that are reverse scored
#' @param rotation what rotation to use c("none", "varimax", "oblimin","promax")
#' @param seed numeric value for setting a seed. Allows results to be reproduced.
#' @param true_model If TRUE, Estimates model using \code{nfactors} only. Is used in \code{enumRsimulation} function. Default is FALSE.
#' @param use If doing covariances or Pearson R, should we use "pairwise" or "complete cases"
#' @param f1_items Total number of items on factor 1
#' @param f2_items Total number of items on factor 2
#' @param f3_items Total number of items on factor 3
#' @param f4_items Total number of items on factor 4
#' @param f5_items Total number of items on factor 5
#' @param f6_items Total number of items on factor 6
#' @param f7_items Total number of items on factor 7
#' @param f8_items Total number of items on factor 8
#' @param f9_items Total number of items on factor 9
#' @param f10_items Total number of items on factor 10
#' @param f1_itemsR Number of reverse-scored items on factor 1
#' @param f2_itemsR Number of reverse-scored items on factor 2
#' @param f3_itemsR Number of reverse-scored items on factor 3
#' @param f4_itemsR Number of reverse-scored items on factor 4
#' @param f5_itemsR Number of reverse-scored items on factor 5
#' @param f6_itemsR Number of reverse-scored items on factor 6
#' @param f7_itemsR Number of reverse-scored items on factor 7
#' @param f8_itemsR Number of reverse-scored items on factor 8
#' @param f9_itemsR Number of reverse-scored items on factor 9
#' @param f10_itemsR Number of reverse-scored items on factor 10
#' ... parameters to pass to the factor analysis program The most important of these is if using a correlation matrix is covmat= xx
#' @return dataframe
enumRsimulation<-function(seed=12345,
ndatasets=200,
patternmatrix=NULL,
effectmatrix=NULL,
nfactors =5,
loading=.5,
items=NULL,
items_p_f=5,
itemsR_p_f=2,
itemsR=NULL,
loading_norm=FALSE,
loading_norm_sd=.05,
# effects matrix
rfactors=0,
r_norm =FALSE,
r_norm_SD =.015,
samplesize=300,
method="ml",
rotation="oblimin",
custom_item=FALSE,
f1_items=NULL,
f2_items = NULL,
f3_items = NULL,
f4_items = NULL,
f5_items = NULL,
f6_items = NULL,
f7_items = NULL,
f8_items = NULL,
f9_items = NULL,
f10_items = NULL,
f1_itemsR= NULL,
f2_itemsR = NULL,
f3_itemsR = NULL,
f4_itemsR = NULL,
f5_itemsR = NULL,
f6_itemsR = NULL,
f7_itemsR = NULL,
f8_itemsR = NULL,
f9_itemsR = NULL,
f10_itemsR = NULL
){
###prep
set.seed(seed)
require(psych)
require(GPArotation)
require(GAIPE)
simulation<-ndatasets
# cleanup items
if(is.null(items)){
if(custom_item){
items=c(f1_items,
f2_items,
f3_items,
f4_items,
f5_items,
f6_items,
f7_items,
f8_items,
f9_items,
f10_items)
itemsR=c(f1_itemsR,
f2_itemsR,
f3_itemsR,
f4_itemsR,
f5_itemsR,
f6_itemsR,
f7_itemsR,
f8_itemsR,
f9_itemsR,
f10_itemsR)
}else{
items=rep(items_p_f,nfactors)
itemsR=rep(itemsR_p_f,nfactors)
}}
## Results
meganames<-c( "dof","chisq","prob","sqresid","fit", "RMSEA","RMSEA_lower","RMSEA_upper","BIC","SABIC","null.model","null.dof","null.chisq","complex","eChisq","SRMR","eCRMS","eBIC","TLI","fa.fit","fa.fit.off","objective","ENull","eSABIC","null.chisq1" , "TLI.m","NFI.m","CFI.m","RMSEA_lower.m","RMSEA_upper.m","eRMS","cfit.1","cfit.2","cfit.3","cfit.4","cfit.5","cfit.6","cfit.7","cfit.8","cfit.9","cresidual.1","cresidual.2","cresidual.3","cresidual.4","cresidual.5","cresidual.6","cresidual.7","cresidual.8","cresidual.9","map.values","rfactors","samplesize","loadings","dataset","error","factor")
mega<-data.frame(matrix(rep(0,length(meganames)),nrow=1))
names(mega)=meganames
mega_null<-mega
mega_null$error<- 1
percentiles<-data.frame(c(1))
names(percentiles)<-"ds"
if(is.null(patternmatrix)){
fmodel=GenFactorMatrix(nfactors=nfactors, items= items[1:nfactors], itemsR=itemsR[1:nfactors], loading=loading,loading_norm=loading_norm,loading_norm_sd=loading_norm_sd)
}else{fmodel=patternmatrix}
if(is.null(effectmatrix)){
smodel=GenStructure(nfactors=nfactors,rfactors=rfactors,r_norm=r_norm,r_norm_sd=r_norm_sd)
}else{smodel=effectmatrix}
## Gen Data
for(d in 1:simulation) {
data=GenData(fmodel=fmodel,effect=smodel,samplesize=samplesize)
cov<-cov(data)
test<- try(results<- enumR(cov,n=nfactors,samplesize=samplesize,covar=TRUE,truemodel = TRUE,fm = method), silent=TRUE)
if ('try-error' %in% class(test)) {mega[d,]<- mega_null
} else {mega[d,]<- results
mega$factor[d]<-nfactors
mega$error[d]<- 0
}
mega$dataset[d]<-d
print(d)
}
mega$rfactors<-rfactors
mega$samplesize<-samplesize
mega$loadings<-loading
ds_true<-mega[mega$factor==nfactors,]
percentiles$RMSEA_.95<-quantile(ds_true$RMSEA,.95)
percentiles$MAP_.95<-quantile(ds_true$map.values,.95)
#percentiles$chisq_.95<-quantile(ds_true$chisq,.95)
#percentiles$BIC_.95<-quantile(ds_true$BIC,.95)
percentiles$SRMR_.95<-quantile(ds_true$SRMR,.95)
#percentiles$SABIC_.95<-quantile(ds_true$SABIC,.95)
#percentiles$TLI_.95<-quantile(ds_true$TLI,.95)
percentiles$chisq_sig_.95<-quantile(ds_true$prob,.95)
#percentiles$RMSEA_.05<-quantile(ds_true$RMSEA,.05)
#percentiles$MAP_.05<-quantile(ds_true$map.values,.05)
#percentiles$chisq_.05<-quantile(ds_true$chisq,.05)
#percentiles$BIC_.05<-quantile(ds_true$BIC,.05)
#percentiles$SRMR_.05<-quantile(ds_true$SRMR,.05)
#percentiles$SABIC_.05<-quantile(ds_true$SABIC,.05)
percentiles$TLI_.05<-quantile(ds_true$TLI,.05)
percentiles$chisq_sig_.05<-quantile(ds_true$prob,.05)
return(list(percentiles,mega))
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.