17 Dimensionality assessment of ordered polytomous items with parallel analysis (Present by Chenwei)

Nicky's

Nicky's

by LI XIAOMIN -
Number of replies: 1
Parallel analysis is a popularly recommended method for assessing the dimensionality of a variable set. This study discussed some aspects needed consideration when selecting an appropriate PA procedure. These aspects include the PA methods (Horn’s PA, PA-PAFA, and PA-MRFA), correlation (Pearson correlation, and polychoric correlation), and the sampling technique (multivariate normal distribution, and permutation). Data for simulations were generated based on a linear CF model, ordered polytomous responses, with both major and minor factors. Simulation results recommended the polychoric 95% threshold PA-MRFA method as the best choice. If non-convergence, the Pearson mean threshold PA-MRFA is another good choice. Polychoric 95% threshold Horn’s PA is also good choice for empirical practice.

Q:
1. What is the “reduced correlation matrix” used in the PA-PAFA?
2. Selection of the PA procedure is related to the type of data. In this study, although polychoric PA-MRFA was recommended, when data type change, how to determine which PA procedure should be used?
In reply to LI XIAOMIN

Re: Nicky's

by LIU CHEN WEI -
1. it is known in linear factor model. The correlation matrix minus unique factor.
2. Didn't get u