xtpcse AR(1) produces high R-squared

[Lilly]

I estimated a model using xtpcse with independent errors within panels and I got R squared of around 30%. When I estimated the same model with AR(1) or PSAR(1) options the R squared jumps to 97%. Can you explain this significant jump in R-squared?

[Muzna Rahmna]

I am facing similar issue. Did you find the reason behind this?