META-ANALYSIS ON PARSIMONY-PARAMETER MODEL FOR EVIDENTIAL ACCURACY AND PRECISION

Abstract

Psychological measurement focuses on the nearest approximation of a construct or latent variable it purports to measure. But being inappropriately parsimonious would constitute failure to model such psychological phenomena and, on the other hand, excessive modeling of phenomena would amount to over-parameterization. This challenge requires suitable psychometric strategies to ensure evidential accuracy and precision. Therefore, a systematic review and meta-analysis, based on two one-sided tests (TOST) using Neyman-Pearson analysis and I-squared (I²) statistic, was conducted towards developing parsimony-parameter model for evidential accuracy and precision. The initial candidate studies generated were 112 but only 15 studies which satisfied the required inclusion-exclusion criteria were retained. TOST for equivalence test, as computed effect size, facilitated determining inference(s) while Neyman-Pearson analysis pre-specified type 1 error and then I² have checked for heterogeneity. The findings [0.9 and 1.1 (H₀: P₁/P₂ < 0.9 or P₁/P₂ > 1.1 versus H₁: 0.9 < P₁/P₂ < 1.1)] rejected the resulting effect size larger than any equivalence bounds which pre-specified type 1 error rate and those showing the true effects that are as extreme as such equivalence. By indicating a balance between parsimony-parameter versus evidential accuracy and precision, this meta-analysis significantly supports the assumption that a parsimony-parameter model facilitates ensuring significant psychometric evidential accuracy and precision. Since the originating equation reflects the features of evidential accuracy and precision it purports to represent, any inference from the resulting model applies to overall psychometric properties and not only to itself. Therefore, the parsimony-parameter model is considered a significant match for determining psychometric soundness which lends credence to measurement quality.

Keywords: Accuracy; Model; Parsimony-Parameter; Precision; Psychometrics.