clusterItems: Item cluster approach to partial MI

In Dani-Schulze/measurementInvariance: (Partial) measurement invariance analysis

Description

Find subsets of items (clusters) which function homogeneously and may be possible anchor candidates. Can currently deal with two-group models for metric items or dichotomous items (Rasch, 2PL, or probit model).

Usage

res_clusterItems <- clusterItems(res_testMI = NULL,
                                        clusterWhat = NULL,
                                        method = NULL,
                                        alphaValue = NULL,
                                        loadThreshold = NULL,
                                        intThreshold = NULL,
                                        ...)

Arguments

res_testMI	Result object of testMI. Alternatively, a MI model can be described using the arguments of testMI (see ...).
clusterWhat	String, either "loadings", "difficulties", or c("loadings", "difficulties"). Indicates, for which parameter type MI does not hold. Usually, this will be indicated per factor in a single string (e.g. "Factor1 configural Factor2 weak")
method	Criteria used for clustering. One can either specify the maximum difference in item parameters between any two items within a cluster "threshold" or base clustering on nonsignificance of item parameter differences of any two items within a cluster "sigTest". The threshold approach is based upon the approach described in Pohl, Schulze, & Stets (in press); Pohl & Schulze (2020) and Schulze & Pohl (2021). The significance test is used as described in Bechger & Maris (2015).
alphaValue	Type I error used for clustering via significance testing. This argument only needs to be specified if method is "sigtest".
loadThreshold	numeric value for the loading threshold. Only used if method is "threshold" and if applicable in the model.
intThreshold	numeric value for the intercept/difficulty threshold. Only used if method is "threshold".
...	Arguments of testMI, if testMI has not been called before and a MI model is to be described.

Value

A list containing

• Data (filtered for potential missings in the covariate)

• Model specification(s)

• Clustering results organized per factor. If inherited from testMI, MI tests are also present.

Usually accessed via summary().

References

Bechger, T. M., & Maris, G. (2015). A statistical test for differential item pair functioning. Psychometrika, 80(2), 317-340.

Pohl, S., Schulze, D., & Stets, E. (in press). Partial measurement invariance: Extending and evaluating the cluster approach for identifying anchor items. Applied Psychological Assessment.

Pohl, S., & Schulze, D. (2020). Assessing group comparisons or change over time under measurement non-invariance: The cluster approach for nonuniform DIF. Psychological Test and Assessment Modeling, 62(2), 281-303.

Schulze, D., & Pohl, S. (2021). Finding clusters of measurement invariant items for continuous covariates. Structural Equation Modeling: A Multidisciplinary Journal, 28(2), 219-228.

Dani-Schulze/measurementInvariance documentation built on Jan. 28, 2022, 1:56 a.m.