Proc Mixed effect interpretation


I am analyzing an experiment in which we tested the effect of different treatments on trauma. So, for example, blood pressure (and other parameters) is followed in animals exposed to trauma and treated in different ways. I am using the Proc mixed procedure to analyze my data with timepoint, treatment and treatment*timepoint as fixed effects. I then look at the lsmeans differences with tukey adjustments to look at individual timepoints.

For some of the parameters, there is a significant treatment or interaction effect, however, when I look at the lsmeans differences, these are not significant after Tukey adjustments.

How do I interpret this result? Are the treatments not different if there are no individual timepoint differences? For many of the parameters I do not see differences in the data when I graph it (the lines for the individual groups are not separated), while for some parameters he groups look very different, although the individual comparisons are not significant. If there is a significant effect, but no obvious pattern in the data and no significant individual timepoint differences, what does that tell me?

I am wondering whether what I see is a result of the many timepoints in my experiment (20), which results in many multiple comparisons, since SAS compares every group at every timepoint - would a different post hoc test be more appropriate?

I am fairly new to this analysis and I have searched the internet up and down, but the only examples I find are all clear cut (significant effect with significant differences). If anyone can help me understand, or knows a paper that gives an example of how to interprete and write up such a result that would be very much appreciated!


Less is more. Stay pure. Stay poor.
Can you describe the study a little bit more. It seems you are looking at repeated BPs in animals that your tortured or didn't torture and you have 20 collected BP measurements. How many measurements before your tortured the animals, and did you randomize which animals would be tortured?

Now you are saying that for example, BP, you have a significant torture * time interaction term. When you are talking about the trends, are you talking about making a spaghetti graph to look at the differences per animal and torture group over time? Also, how many animals were there?