Help! Data layout and which test?

#1
Hi, :wave:

I have some data on a clinical trial. 2 Independent groups, running in paralell, treatment and placebo(n=50).

Biological markers were determined at the start of the study and at the end. Am trying to find out if there is an effect of treatment? :(


Treatment (start - end)
Placebo (start - end)

I assumed this was a 2x2 mixed design anova (repeated measures). I am not sure how to arrange this data (have tried in SPSS but am not given Tukey values so assume either incorrect test or incorrect layout of data.

If anyone can help with the layout and test I would be extremely grateful!

Thanks :)
 
#2
Yes, you're right. It's a 2x2 mixed design anova.

If you're running it using repeated measures in GLM, data for start and end should be in two columns. A third column would indicate treatment or placebo--that one would be between subjects.

You won't get Tukey results since you have only two levels of each factor. You need at least 3 for Tukey. Just use the p-value from your F test for each factor to determine if the two means are different.

Karen
 
#3
Mixed ANOVA or Linear mixed model ??

Hi,

Experimental data: 2 different, uneven groups of subjects, receiving either treatment A or treatment B, with a measure taken at baseline and another a few months later.)

I carried out analysis on the above data set as a 2x2 mixed design anova for my experiement below - however I am now concerned this should be a linear mixed model following discussion with colleague.

Could anybody shed any light on which instances the two should be used? I understand the concept of the 2x2 mixed ANOVA to give group effect, treatment effect and interaction effect but do not understand the added benefit of the linear mixed model (and changing the data to a 'long' dataset) which has been suggested to me.

All advice greatly appreciated.
 
#4
A mixed ANOVA is just one kind of linear mixed model, which can do either mixed regression or ANOVA.

In most statistical software, you can run a mixed ANOVA model using either repeated measures ANOVA procedure, with data in the wide format or using linear mixed model procedure, with data in the long format.

The former suffices for simple designs, the latter is necessary for more complicated ones. It sounds like yours is pretty simple, but perhaps your colleague sees something in it that warrants the more complicated analysis.

It is absolutely analogous to using an ANOVA procedure vs. a GLM.

Karen