Exercise

# EFA with Principal Axis Factoring

Let's look at another popular extraction method, `Principal Axis Factoring`

(PAF). `PAF`

's main idea is that communality has a central role in extracting factors, since it can be interpreted as a measure of an item’s relation to all other items. An iterative approach is adopted. Initially, an estimate of the common variance is given in which the communalities are less than 1. After replacing the main diagonal of the correlation matrix (which usually consists of ones) with these estimates of the communalities, the new correlation matrix is updated and further replacements are repeated based on the new communalities until a number of iterations is reached or the communalities converge to a point that there is too little difference between two consecutive communalities.

Instructions

**100 XP**

- Create an
`EFA`

model with 4 factors based on`hsq_polychoric`

, but this time use the principal axis factoring method,`pa`

. - Sort variables' communalities in
`hsq_correl_pa`

, from highest to lowest to get a better idea, as to which variables load well on the factors we chose. Recall that you need to use the appropriate attribute of the`hsq_correl_pa`

object, shown in the video. - Now, sort variables' uniqueness in
`hsq_correl_pa`

in the same way.