Mixed model Vs repeated measure Anova

fourize

New Member
#1
Hello,

I have two questions that I would like to clarify, please.
1. What is the difference between a repeated measure ANOVA and a mixed model?
2. Can we use a mixed model when we have only two repeats? for example only 2 visits.

Thank you in advance
 

fourize

New Member
#3
Hi Fed2,

Thank you for your answers. Let take two examples for more understanding.

Suppose I have a continuous variable (CV) to explain.
If I have on the one hand: (1)
CV measures and an other categorical variable at different visits

and on the other hand: (2)
CV measures and other explanatory variables (categorical and continuous) at different visits.

Can I say that in case (1), a repeated measures ANOVA is the best method and in case (2) a mixed model?

Thanks
 

fed2

Active Member
#4
ummmmmm.....

Repeated Measures ANOVA is usually one of two varieties, 1-way Repeated Measures ANOVA and 2-way Repeated Measures ANOVA.

2-way Repeated Measures is more common. In such a Case you would have two explanatory variables, one (usually time) which is 'within subjects', ie each subject experiences all time pionts, and one which is 'between subjects'.

Its hard for me to tell whether that fits
If I have on the one hand: (1)
CV measures and an other categorical variable at different visits
Maybe you can tell me?

Anyway I wouldn't over-think it, the answer is basically always 'mixed model'. RM-ANOVA is really only important in that it is computationally simple, ie it does not require iterative algorithms, but in modern times this is not verry important.
 

katxt

Active Member
#7
No? Obviously I have been under the delusion that a GLM with a nested factor was a mixed model. We live and learn.
 
#10
I'm currently writing a SAP where I'll add a mixed model. Do you guys have a good online reference document on this subject ?
I mean, where I can see how to perform this model using SAS and the points I need to pay attention to!

Thank you a lot
 

fed2

Active Member
#11
SAS Proc mixed. Use the SAS documentation. There are some good we sites if you just google 'proc mixed repeated measures' that show the codes. Fair warning: Sometimes in SAP type situatuions, it is better to run the t-test on the change from baseline scores at select time-points. This is often favorable from a transparency standpoint, although formaly a t-test is a linear mixed model, and can be run with proc mixed. Good luck.
 

hlsmith

Less is more. Stay pure. Stay poor.
#12
Multilevel model are typically preferred, for one reason they can handle missing data. There is a multilevel models in SAS book by I think Wang that is good. But as @fed2 mentioned, SAS documentation may be sufficient. One thing many say to use MLM one should have quite a few groups!