I’m wondering how to report the results from the new linear mixed modelling analysis? I’m very familiar with GLMMs for ecological data, where we report effect size/estimate/beta, confidence intervals, R2 including random effect. But after running the mixed model here the output is a series of column names I don’t know and there is no legend.

I haven’t been able to find a paper that uses this analysis and reports these results. I’m hoping I don’t have to do this first stage in metaboanalyst and rerun the models in R for each significant feature to generate the appropriate results…

At the top of the page, we say that the modelling is using the limma R package and link their excellent user manual.

For models that include a blocking factor, we are following the methods specified in section 9.7 almost exactly. The only difference is that in the final line of section 9.7, they specify a single coefficient from the model (referencing a single contrast) in their topTable command, whereas we do not specify any which causes topTable to report all coefficient values for the primary metadata

When the primary metadata is continuous (ie. AGE) or categorical with only two groups (ie. CTRL, TRT), there is only one coefficient (AGE or TRT-CTRL respectively) and limma returns the moderated t-statisic and associated p-value; when it is categorical with >2 groups there are more coefficients (ie. CTRL, TRT.A, TRT.B = TRT.A-CTRL, TRT.B-CTRL). When topTable is not given a specific model coefficient, it reports an ANOVA-like F-statistic and associated p-value. The MetaboAnalyst results table column will be named either “F” or “t” depending on the situation.

If you’d like to dig further, query the “topTable” documentation page in R (?topTable; limma R package). They have an excellent description of each column. To see the exactly how the code is implemented, you can click “Show R Commands” in the top right corner and then inspect the relevant functions in the MetaboAnalystR github page.

Thanks for the response, I did have a browse of the limma user manual but thank you for highlighting the relevant sections.