Dear all,
I have tried to find the answer from previous posts, but I couldn't find any clear explanation. My question is regarding paired analysis in limma. If I want to compare weight changed before and after a diet, can I use weight a continuous variable?
For instance if 100 individuals go on a diet, we measure their weight before and after the diet. (Here I am assuming that all of them lost weight after diet.) As each person can loose different kgs, I would imagine continuous variable is more appropriate, rather than a categorical variable .
This is my design matirx for paired analysis comparing same individual before and after diet: design <- model.matrix(~Subject+weight) # here Subject is the person ID fit <- lmFit(input,design) fit2 <- eBayes(fit) result <- topTable(fit2,coef="weight",adjust.method="bonferroni")
I was wondering if this analysis is any different from the categorical analysis, where all the individuals are marked heavy (before diet ) and lean (after diet). For which the design would be design <- model.matrix(~Subject+status) # here status is heavy or lean.
Thanks for your help in advance.
Regards, Spurthy.
•
written
by
spurthy111 •
Yes, this is a continuous variable in the model, though in practice you might get clearer results with the categorical "status" method (give them both a try and see which seems to be more informative).
The only thing to keep in mind with continuous variables is that the coefficients are in terms of kg or whatever else you measure in. So if someone dropped 10kg and had a 2x change in gene expression then the per-kg fold change would be...rather less dramatic. Keep that in mind when looking at the results.
Thanks a lot Devon! I have tried both continuous and categorical "status" methods. Limma gives better results for continuous variable method. And thanks for reminding about the coefficients.
Log in to answer this question.
Wt loss would be discrete variable (IMHO) assuming that wt loss is calculated in kg.