Regular Learner
Posts: 1

# Stopped because of infinite likelihood - Not due to duplicate records

Hello community,

I am receiving the "stopped because of infinite likelihood" error message and could use your help resolving this issue. I have read other posts on this board and one common suggestion is to check that there are no duplicate entries for a subject. I have ensured that each subject has only one value per time point. I am hoping you all might be able to help me figure out what else could be causing this error. Below is an explanation of my design and some samples of my data and code. Anyone advice you could give would be greatly appreciated. Thank you!

To give you some more background on my design and code:

I am running a 3-way ANOVA (Age | Diet | Time) with repeated measures on Time using Proc Mixed. Each subject, is assigned to one age (2 levels) and one diet (2 levels) and have one value per time point (223 sampling times for each subject). Therefore, there are 223 values per subject and a total of 33 subjects (=7392 total lines of data).

Here is a sampling of how my data is entered:

input slice\$ age\$ diet\$ time slope;

cards;
1a Young Control 0.0 106.1145644
1a Young Control 0.3 95.66878936
1a Young Control 0.7 96.04325189
1a Young Control 1.0 95.74211497
1a Young Control 1.3 95.23252855
1a Young Control 1.7 89.55785791
1a Young Control 2.0 98.95222987

And the code for my model:

proc mixed data= AgedRepopLTP;
class slice age diet time;
model slope = age|diet|time / residual;
repeated time / subject=slice type=ar(1);
random slice;
run;
run;quit;

Posts: 2,655

## Re: Stopped because of infinite likelihood - Not due to duplicate records

While I am a big fan of the AR+RE method proposed by Littell, and have recommended it a lot, I suspect that is what is causing your problem.  The first thing to try is to remove slice as a random effect.  This is the equivalent of assuming that all subjects have a common intercept in the mixed model equations, which does not seem unreasonable from the smattering of data provided.

If that doesn't solve things, you may need to look at more flexible covariance structures than AR(1)--you may have data that makes that assumption about the errors run into difficulties.  With that many time points, a spline structure may be able to give you what you need.

Steve Denham

Discussion stats