[R-sig-ME] tracePlot for a huge MCMCglmm model

[dani]

Hello everyone,


I obtained a very large MCMCglmm object and I tried to plot it. I have about 7000 observations nested into two groups of about 1500 each. Just to give you an idea, if I attempt to save the model as a rds, I succeed, but I cannot read it afterwards because of memory allocation issues. My issue is that I cannot plot it nor perform a diagnostic because of its size.


My questions are:

Should I conduct my analysis on a smaller random dataset (20% of the data or 50% of the data) and present those results graphically and conduct a diagnostic test for those? Is it ok to report those results and make comments about the potential convergence of the model based on the full dataset? My instinct tells me it is probably not ok to do this, but I thought I should ask about this:)


Thank you very much!

Best,

Dani

	[[alternative HTML version deleted]]