PCA over the panel dataset

[Zhijian Zhou]

I am a undergraduate student working on a econometrics research. The problem is here. I have gotta panel data of several indicators from more than 3000 public companies during an interval from 2004 to 2018. I would like to consolidate these indicators to one, in another word, I would like to calculate a score for each company in each year based on these indicators(Suppose that I have selected the right indicators.) and then use this score as explained variables in my next step of regression.The first thing come into my mind is the pca method. But usually it is not appropriate for panel data. Some information I got from others says that it also incurs some problem if I choose to use pca one time a year.(doing the regression for 15 times and calculate a score measure each year and combine them). Also I heard about gpca may be applicable for panel dataset, but no clue how to practice with Stata.
Very appreciated if there is any advice!

[Joseph Coveney]

I don't know what gpca is (generalized principal component analysis?), but maybe you could use gsem and model the indicators as loading onto a factor (latent variable) or a limited set of factors. gsem can model repeated measurements on the set of public companies, as well.

[Zhijian Zhou]

Yes, gpca refers to generalized principal component analysis.
Appreciated for your great idea! I have looked into gsem method, but is it normal for it to take more than a day to run this gsem model linking to fixed-effect model? I am just uncertain about the remaining time to finish it and a little bit anxious.