PCA for repeated measures

[Jean-Michel Galarneau]

Hello Statalist contributors,

I am looking for a solution to a problem that seems unanswered on this forum as of now even if there are people asking similar questions.

I have 42 blood biomarkers and it is common practice to reduce those to a smaller number of factors to understand the data. I therefore would conduct a PCA on it and go from there but in this particular case, I have two sets of these data on the same participants with 3 months apart. I would like to compare PCA's but of course I cannot simply conduct a PCA at each time point and compare them so the question is, are there known methods for conducting and comparing PCA's from two time points? I have found that there is a program called Multiple Factorial Analysis in R and an article on the subject but sadly I am simply inept in R. Is there something equivalent in Stata?

Any advice would be greatly appreciated.

J-M

[Gobinda Natak]

I'm clearly not an expert in this, but

I would like to compare PCA's but of course I cannot simply conduct a PCA at each time point and compare them

why not? Ideally the results are the same.

More formally, you could do a PCA for the first wave and then do a confirmatory factor analysis with the -sem- commands for the second time point.

[Jean-Michel Galarneau]

The idea is to compare factor1 at time point 1 to factor 1 at time point 2 but of course these will not load on the same biomarkers from time 1 to time 2.

[Dave Airey]

If you were interested in a PCA that reflects the change from time1 to time2, why not get a summary statistic for that like the difference or AUC?

Here are some references:

https://www.sciencedirect.com/scienc...67811602000654
https://personal.utdallas.edu/~herve...S-mfa-2013.pdf

[Jean-Michel Galarneau]

The second paper I have read and is actually what I am sort of referencing in my original question and I believe it fits what I am trying to do but was hoping I would not have to use R.