# Obtaining individual factor loadings from a CFA in an experience sampling design

I have these intensive longitudinal data of n = 300 participants with 55 measurement occasions each. I have 12 emotion items (6 positive and 6 negative, both high and low arousal).

I am interested in emotion differentiation. People that are able to precisely describe and distinguish between different emotional states generally score well on measures of general well-being. I want to examine whether emotion differentiation differs as a function of the situation: specifically the dichotomy of social versus non-social situations.

I wanted to perform a factor analysis with two factors: positive affect and negative affect. Calculate unique loadings per person and use its variance to operationalize emotion differentiation per factor.

1) I would like to know if there is way to run a confirmatory factor analysis, let the factor loadings vary across individuals, and obtain the loadings per individual. I have Mplus and R lavaan installed.

2) I also wonder if the number of occasions is enough (55). Wouldn't I have a lack of power?

I am open to all suggestions!

• Can you expand on what you're trying to do, and why? Power to test what null hypothesis? Apr 21, 2018 at 20:30
• Before proceeding to a classification protocol (e.g., separating the respondents two categories), I would suggest a multilevel CFA to confirm that there is some measurable variation in the parameter estimates for the individuals. And if so, on what manifest variables the most variation is being observed. This is possible using Mplus, but not lavaan (check the manual for multilevel CFA). Apr 23, 2018 at 13:08
• @GreggH. Thanks. Yes that would make sense. Different papers have found substantial variance within persons in different affect items. Probably it is possible in R though, I think. Huang, F. L. (2017). Conducting multilevel confirmatory factor analysis using R. If I follow these procedures that is, based on Hox (2010): Multilevel Analysis Apr 23, 2018 at 13:13
• I'm very interested in the Huang 2017 reference...¿is it using sem or lavaan? Apr 23, 2018 at 13:17
• They use lavaan , but also provide an additional function mcfa.input() . All the steps and scripts, with example data, are described in the article. See: faculty.missouri.edu/huangf/data/mcfa/MCFAinRHUANG.pdf Apr 23, 2018 at 13:23