Item parcels in structural equation modelling: an applied study in sport management

Abstract

The purpose of this study is to describe the influences of different parceling strategies on goodness-of-fit measures and parameter estimates of a sport management structural model with latent variables. The use of small sample sizes to test models with a large number of parameters can produce poor fit indexes, mainly because many indicators tend to increase the chances of cross-loadings, which in turn reduce the common variance. Considering that in social and behavioral sciences is not quite easy to have access to large-enough samples, item parceling has been proposed as a remedy for this kind of situation. Using a theoretically-supported sport management model and real data, we compared total disaggregation model (items as indicators) with partial disaggregation models (parcels as indicators) and total aggregation model (summated score as the indicator). Results showed that different strategies of parceling could lead to very distinct conclusions. Implications for future studies using parceling in the field of sport management are discussed.