View in EDS

Exploring the Effect of Item Parceling Strategies and Number of Items per Parcel on Measurement Invariance Testing in Confirmatory Factor Analysis

Saved in:
Bibliographic Details
Title: Exploring the Effect of Item Parceling Strategies and Number of Items per Parcel on Measurement Invariance Testing in Confirmatory Factor Analysis
Authors: Cao, Chunhua, Liang, Xinya
Source: Tutorials in Quantitative Methods for Psychology, Vol 21, Iss 2, Pp 95-107 (2025)
Publisher Information: The Quantitative Methods for Psychology, 2025.
Publication Year: 2025
Subject Terms: measurement invariance, distributed parceling, item parceling, isolated parceling, Psychology, fit measures, BF1-990
Description: Item parceling is common in structural equation modeling used in psychology, especially when the research interest is in the structural relationship between latent factors, but research on its performance in measurement invariance testing is scarce. This study examined the sensitivity of fit measures (LRT, RMSEA, CFI, AIC, BIC, and SaBIC) to measurement noninvariance with different numbers of items per parcel and parceling strategies using a simulation study. The design factors included the location of measurement noninvariance, the magnitude of measurement noninvariance, the proportion of items with measurement noninvariance, sample size, number of items per parcel, and parceling strategies. Results suggested that compared to no parceling, isolated parceling tended to increase the sensitivity of fit measures, especially when fewer parcels with more items per parcel were used; distributed parceling showed lower sensitivity, particularly when noninvariant items were evenly distributed and fewer parcels were used. The findings provide insights into the use parceling in examining the relationship between constructs. The paper concludes with practical implications and future research directions.
Document Type: Article
ISSN: 2292-1354
DOI: 10.20982/tqmp.21.2.p095
Access URL: https://doaj.org/article/674f2b8ef4a5461283a9efe46f468d80
Accession Number: edsair.doi.dedup.....42150bff29dd930cfab9d38f50483cdc
Database: OpenAIRE
Description
Abstract:Item parceling is common in structural equation modeling used in psychology, especially when the research interest is in the structural relationship between latent factors, but research on its performance in measurement invariance testing is scarce. This study examined the sensitivity of fit measures (LRT, RMSEA, CFI, AIC, BIC, and SaBIC) to measurement noninvariance with different numbers of items per parcel and parceling strategies using a simulation study. The design factors included the location of measurement noninvariance, the magnitude of measurement noninvariance, the proportion of items with measurement noninvariance, sample size, number of items per parcel, and parceling strategies. Results suggested that compared to no parceling, isolated parceling tended to increase the sensitivity of fit measures, especially when fewer parcels with more items per parcel were used; distributed parceling showed lower sensitivity, particularly when noninvariant items were evenly distributed and fewer parcels were used. The findings provide insights into the use parceling in examining the relationship between constructs. The paper concludes with practical implications and future research directions.
ISSN:22921354
DOI:10.20982/tqmp.21.2.p095