Measurement reliability, construct validity, and transparent reporting in original and replication psychological research.

Abstract

Published (replication) studies that use measures to assess psychological phenomena need to transparently report information on measurement procedures, validity, and reliability to allow verification and use in future (replication) research. However, earlier results highlighted widespread poor reporting of psychological measurement. Here, we investigated measurement reporting in a sample of 77 measures within 56 Many Labs replications and related original articles (14-17) and found that the information relevant for reusing measures was reported in full in around half the replication measures, and only in 5.2% of the original studies. We also observed that around a third of multiple-item measures in original studies and 11.4% in replications reported reliability coefficients, with comparable proportions for reporting any convergent, discriminant, predictive, or factorial validity evidence. We assessed the reliability, unidimensionality, and measurement invariance of multiple-item measures using the openly available Many Labs item response data. We observed that while some measures passed these psychometric checks, they rarely did so consistently across labs. These results corroborate existing findings that measurement reporting in published research lacks transparency, and that poor measurement reporting may obscure insufficient reliability and validity. We offer suggestions on how to improve measurement reporting practices and increase the use of validated measures.