This paper introduces a family of goodness-of-fit measures for partial least squares path modeling (PLS). It shows that PLS estimates the parameters of a composite factor model. The discrepancy between the empirical covariance matrix and the covariance matrix implied by the composite factor model permits quantifying the model fit. The paper proposes statistical inference based on the bootstrap as an exact test of model fit. Moreover, it suggests the standardized root mean square residual as an approximate measure of model fit. These two goodness-of-fit measures enable analysts to empirically assess the validity of interrelated composites of observed variables, i.e., conduct a confirmatory composite analysis by means of PLS. Finally, the paper provides recommendations for conducting and reporting typical PLS analyses.
|Publication status||Published - 9 May 2014|
|Event||The 3rd Annual Conference of the Dutch/Flemish Classification Society (Vereniging voor Ordinatie en Classificatie - VOC) - Leiden, Netherlands|
Duration: 9 May 2014 → 9 May 2014
Conference number: 3
|Conference||The 3rd Annual Conference of the Dutch/Flemish Classification Society (Vereniging voor Ordinatie en Classificatie - VOC)|
|Abbreviated title||3rd VOC Conference|
|Period||9/05/14 → 9/05/14|