Some data does not fit a t-test, and treating it as though it does is the quiet source of a lot of irreproducible results.

We build models for data with structure: linear mixed-effects models for repeated measures, biomarker association in both cross-sectional and longitudinal studies, and hidden Markov models for sequential data. We also model mouse behavior to characterize learning and memory.

This is the Core's oldest strength and the reason a study design conversation early is worth so much. If your measurements are nested, crossed, correlated over time, or collected in batches, the modeling choice is usually more consequential than the pipeline choice.