Dimensionality Analyses

Jana Welling, Christian Diegel

2026-09-23

General Information

The dimensionality of a test indicates how many latent dimensions are required to answer the test. It affects the IRT models used. For example, a key assumption of the Rasch model is that unidimensionality holds.

To test this assumption, both a unidimensional model and multidimensional models are estimated and compared. Dimensionality variables are item characteristics that allow items to be divided into subgroups. These may include, for example, overarching tasks with multiple items, item type, or— in reading tests—text type. The evaluation is based on analyses of variances and correlations of the dimensions of the assumed multidimensional model. For model comparison, information criteria such as AIC and BIC are again used.

These analyses can take a very long time in R (several hours), so it is recommended to run them on a particularly powerful computer. Multidimensional analyses are especially computationally demanding; therefore, instead of quadrature methods, Monte Carlo methods are used. For trial runs of the syntax, snodes = 10 can be used initially, but for final estimation a higher number should be selected (default snodes = 5000).

The dimensionality analysis should only be conducted for all valid cases (function argument valid). It is typically performed using the full sample rather than separated by groups.

Main Function

In the NEPSroutines package, the function dim_analysis() is used to assess the dimensionality of a test. The function can estimate and compare models for multiple dimensionality variables simultaneously (argument dim). It performs the analyses and returns the required results in the form of a table as well as additional information in the console. In addition, the results are saved as an Excel file and in RDS format. If the file names need to be modified (e.g., because a test is scaled simultaneously for several cohorts), use the function argument name_group that adds a suffix to each file name (e.g., "cohort_6").

In the example, a multidimensional model is estimated for the item type and a multidimensional model is estimated for the text type.

NEPSroutines::dim_analysis(
  resp = resp,
  vars = vars,
  select = "mixed",
  valid = "valid",
  dim = c("kog_type", "text_type"),
  scoring = "scoring",
  irtmodel = "PCM2",
  verbose = TRUE,
  return = FALSE,
  save = TRUE,
  print = TRUE,
  path_table = "tables",
  path_results = "results",
  overwrite = TRUE
)

Information on the individual function arguments can be obtained via the help function (?dim_analysis).

Minor Functions

In addition to the main function, the following minor functions can be used:

Function Input Output
conduct_dim_analysis Data.frames containing information on items and person characteristics, as well as the specification of the items under investigation. Additionally, the IRT model and the scoring information created during data preparation in vars must be provided. Results of the dimensionality analysis (R object, optionally RDS)
dim_summary Output of the function conduct_dim_analysis (or the element analysis from the output of dim_analysis) Tables with results of the dimensionality analysis (R object, optionally Excel)
print_dim_summary Output of the function dim_summary (or the element summary from the output of dim_analysis) Summary of the dimensionality analysis (printed in the console)

Information on the individual function arguments can be obtained via the help function (e.g., ?dim_summary).