Statistical Issues

In going through the analyses of metabolites for each cohort and meta-analyses, we encounter some statistical issues. On this page, we try to describe these and how the app handles them after consultation and consensus with statistical work group.

Cohort-specific analyses

1. Singularity
2. Multiple platforms, same metabolite on a participant
3. Missing values and imputation strategy
4. Low variance metabolites
5. Scale of metabolite values across cohorts (ie transformations)
6. Pooling several nested case control for the same cohort - cohort batch run

Meta-analyses

1. Cohort characteristics
2. Tranformation of association to unify estimates (ie per sd) for all model types
3. Multiplicity adjustment approach