topic Re: MIXED models for within/between-subjects interpretation in Statistical Procedures

MIXED models for within/between-subjects interpretation

Sotch — Mon, 23 Dec 2013 15:54:00 GMT

Hi,

I have continuous data for 8 subjects and around 4-5 data for each subject.

I would like to see the correlation between the dependent variable and the independent variable.

But first, I want to see if I can mix the data together without taking into account that there is 8 subjects.

In fact, I want to see if samples taken from the same donor are as independent as the samples taken from different donors.

Therefore, I have to justify mixing all samples together which would ease my analysis.

Can anyone tell me if it is correct: To justify mixing all samples together, I have to see if the variance within-subject is less than the variance between-subject? Or is it the contrary?

To do that, I used the donor as a random effect in Mixed models procedure and I compare the within-subject effect (residual variance) to the between-subject effect (donor variance).

Thank you very much!

Re: MIXED models for within/between-subjects interpretation

statistician13 — Mon, 23 Dec 2013 16:10:17 GMT

Take a look at the COVTEST statement in proc MIXED.

Re: MIXED models for within/between-subjects interpretation

Sotch — Mon, 23 Dec 2013 16:25:40 GMT

Hi!

I already have my within-subject variance and between-subject variance.

I need help in the interpretation of these results.

My goal is to justify mixing the samples together.

I would like to have your point of view... To justify that I need to prove that the within-subject variance is less than the between-subject variance? Or that the within-subject variance is more than the between-subject variance?

Thank you!