[R-sig-ME] Advice with autoregressive correlation specification in lme()

[Santosh Srinivas]

Dear List members, I am running into an error when specifying autoregressive correlation structure in lme(). Here are the model specifications:

m1 = lme(y ~ time, random = ~1+time|fighter_id, data = df, na.action = na.omit)
m2 = update(m1, correlation=corAR1())

m1 runs fine. But, m2 throws the following error:
Error: 'sumLenSq := sum(table(groups)^2)' = 5.44838e+10 is too large.
 Too large or no groups in your correlation structure?

Not sure what the error is as there are enough groups.
Number of Observations: 3586633
Number of Groups: 1109

What are we doing wrong? How can this error be overcome?

Thanks!
sbs

	[[alternative HTML version deleted]]