LaRA: Latent Regression Analysis

R package for Bayesian estimation of latent trait distributions considering hierarchical structures and partially missing covariate data. Currently LaRA allows the user to fit unidimensional latent regression item response models (LRMs) and their extensions for clustered observations. LRMs apply a multivariate regression equation to model the relationship between the latent trait and additional person covariates. Thus, they combine the fields of measurement models and structural analysis. LRMs are typically employed to generate plausible values in large-scale assessments.


Installing LaRA

To install the latest development version from GitHub using the devtools package, run:



LaRA relies on some routines from other R packages, where the latest CRAN version is in use: mvtnorm, ucminf and rpart.


Imputation of five plausible values with the multigroup dataset:


## prepare data input
Y <- simdata_2mglrm[, grep("Y", names(simdata_2mglrm), value = TRUE)]
X <- simdata_2mglrm[, grep("X", names(simdata_2mglrm), value = TRUE)]

## estimation setup: MCMC chains of length 120 with 20 initial burn-in samples
## for testing purposes (for your applications itermcmc > 10000 needed)
results <- mglrm(Y = Y, X = X, S = simdata_2mglrm$S, itermcmc = 60, burnin = 10, thin = 2)

PVs <- t(results$MCMCdraws$Theta[sample(nrow(results$MCMCdraws$Theta), size = 5), ])


AƟmann, C., Gaasch, J.-C., Pohl, S., & Carstensen, C. H. (2015). Bayesian estimation in irt models with missing values in background variables. Psychological Test and Assessment Modeling, 54(4), 595-618.

jcgaasch/LaRA documentation built on Dec. 20, 2017, 12:19 p.m.