Hi, I just had a quick question on the Feature Details Table found in the linear model option of the MetaboAnalyst Software.
Could you confirm that the “P-value” is the raw P-value and the adjusted P-value is the FDR adjusted P-value? Could you also provide the FDR adjustment technique (Benjamini-Hochberg method?)?
Thanks.
Yes, both of those things are correct. P-value = raw p-value. Adjusted p-value = Benjamini-Hochberg method (as implemented in the limma package).
Thank you! I appreciate the clarity.
Just as a sidenote, part of me wonders if this confuses other people because the figure (that shows which metabolites are impacted by treatment before and after covariate adjustment) in the “linear model with covariate adjustment” page suggests that there are significant treatment effects for my dataset, but then when you look at the FDR adjusted P-values in the feature details table, you realize that there aren’t any significant treatment effects, at least for my case. I could see some people becoming deceived into thinking that there are significant results when those metabolites may not be impacted after FDR adjustment. I guess I view the FDR-adjusted p-value as the “default setting.” Just a comment, not sure what others may think.