Item Details

title

BAYESIAN APPROACHES TO QUANTIFYING THE PRACTICAL IMPACT OF MEASUREMENT NON-INVARIANCE: EXTENDING dMACS

author

Lacey, Conor

abstract

A measurement construct can be defined as the model linking a measure to a latent construct. Measurement invariance assesses the psychometric equivalence of a measurement construct across groups. Whenever a measurement construct is said to be non-invariant this implies the measurement construct has a different structure or meaning across groups and therefore groups cannot be meaningfully compared under the same measure. Different methods exist to detect measurement invariance (e.g., MGCFA, MNFLA, MH, etc.), however, it has become clear with advancing research that complete measurement invariance isn’t an absolute necessity for group comparison (e.g., partial invariance) and there is no clear threshold for non-invariance. As a result, the field has begun to turn to effect-size research for answers. One of the issues with effect size research however is its inability to consider the plausibility of a zero-effect size in the calculation of an estimate. In this study, we take the commonly used measurement invariance effect size estimator d_{MACS} and modify it using a spike-and-slab approach created by van den Bergh et al., (2021) that allows the probability of an effect of zero to be considered in its calculation. We evaluate the implications of this method.

subject

Bayesian

Effect Size

Measurement Invariance

Psychometrics

Spike-and-Slab

contributor

Cole, Veronica (advisor)

Cole, Veronica (committee member)

Furr, Mike (committee member)

Garrison, Mason (committee member)

Bollen, Kenneth (committee member)

date

2023-09-08T08:35:23Z (accessioned)

2023-09-08T08:35:23Z (available)

2023 (issued)

degree

Psychology (discipline)

identifier

http://hdl.handle.net/10339/102612 (uri)

language

en (iso)

publisher

Wake Forest University

type

Thesis