holdout: Holdout Validation

In AnthonyRaborn/cvIRT: Cross Validation and Bootstrap Methods for IRT Model Selection

Description

Holdout Validation

Usage

holdout(responseData, modelTypes, proportion, replications,
  indicator = TRUE, ..., seed = NULL, type = "person")

Arguments

responseData	The initial, full sample data as a matrix of item responses.
modelTypes	A character vector specifying the model types to be compared. Uses the 'TAM' package format, so must be one of the following: "1PL", "2PL", "PCM", "PCM2", "RSM", "GPCM", and "2PL.groups".
proportion	The proportion of the data to hold out for validation. Must be a numeric value between 0 and 1, exclusive.
replications	The number of times to replicate the holdout process using the same data, models, and proportion.
indicator	A logical value that controls the progress printing.
...	Further arguments to be passed to the 'tam' function.
seed	Either a positive integer setting the random seed, or 'NULL'.
type	A character vector specifying whether the validation treats the "person" or the "item" as the unit of observation. Default is "person".

Value

An object of class "cvIRT" with the following values:

seed	The random seed that produced the results.
trainData	A list of each replications' training data.
testData	A list of each replications' testing data.
testLik	A matrix of the loglikelihood values estimated on the testing data for each model within each holdout replication.
nModelParams	A matrix of the number model parameters estimated on the training data within each holdout replication.
AIC	A matrix of the AIC value for each holdout replication, as well as the mean value across each replication.
AICc	A matrix of the AICc value for each holdout replication, as well as the mean value across each replication.
BIC	A matrix of the BIC value for each holdout replication, as well as the mean value across each replication.
-2 log-Likelihood Ratio Test	A list of the log-likelihood ratio test statistics, degrees of freedom, and p-values for each model comparison and each replication, as well as the test using the mean test statistics and degrees of freedom. If only one model is used, returns 'NULL'.
warnings	A character vector of any warnings incurred while the method runs.
time	A vector of the start and end times of the function.

Examples

#None.

AnthonyRaborn/cvIRT documentation built on Jan. 9, 2020, 3:26 a.m.