0
$\begingroup$

I am analysing metabolomics data (~1,000 measurements) across 30 samples from 10 subjects.

The samples are from 2 treatments groups of 5 subjects from whom 3 muscle samples were taken from 3 locations (2 from top and bottom of one muscle (A) and the other from a different muscle B). The initial hypothesis was that the treatment*muscle location effect on metabolite levels. This was to be tested using a repeated measures two-way anova.

However, initial analysis shows that the 2 samples from muscle A differ minimally compared to difference between A vs B. So I am worried in an ANOVA using all 3 locations the two samples from muscle A would function as replicates and this may lead to over estimation of significance of location effect.

I was wondering if my concerns are correct and whether I should just take average from the samples from A and then go with that.

$\endgroup$

1 Answer 1

0
$\begingroup$

If you started with a particular hypothesis and an associated analysis approach, it's usually not a good idea to change them based on a first look at the data.

Your repeated-measures ANOVAs will provide estimates of differences between the two A-muscle locations and of each A-muscle location versus the B muscle. It's possible that the complete analysis will show some metabolites for which the A-muscle location does matter in a practically and statistically significant way. Don't throw away that possibility at the start by throwing away the top vs. bottom A-muscle site information.

I was wondering ... whether I should just take average from the samples from A and then go with that.

That is forcing the A samples to "function like replicates" in a way that separate analysis of the locations wouldn't do. It would seem more likely to over-estimate the "significance of location effect" than continuing with your original analysis approach.

Starting with averages between 2 A-muscle locations will tend to average out technical measurement errors from the A muscle in a way that wouldn't be done for the B muscle. Keeping the A-muscle locations separate in the initial analysis doesn't. If you start with your planned separate analysis of the A-muscle locations you can still choose to compare the average A-muscle values against the B-muscle values in a contrast.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.