Hello,

I'm searching for a Stata-Package or solution that will allow me to factor analyze longitudinal and overdispersed count data, with the overdispersion being my major concern. I have individual-level data (approx. 600 individuals) on a weekly level (52 weeks) with overdispersed counts of 7 different types of behaviors with very low ranges where I observe a maximum count of 20 for one variable and maximum counts of up to 3 for the other ones. However, for most of the observations, the counts are zero or one and the data are, thus, highly skewed.

The problem is, that with standard factor analysis I might violate the assumptions and also, due to the granularity of the data and excess zeros. With standard factor analysis the extracted factors explain a rather low proportion of the variance (smaller than 60%). I already looked into the discussion on the appropriateness of using exploratory factor analysis for binary variables (as the count data rather takes on the form of binary data than continuous ones due to the granularity of the counts) and it violates the assumptions of factor analysis, due to the non-continuous nature.

Right now, my model in this research project is based on factor scores for two factors extracted from a regular factor analysis (command 'factor') with principal component factors (option 'pcf') of the 7 behaviors mentioned above.

I appreciate any support and want to thank you in advance.

All the best,

Welf