Testing for variance effects when the causal genotype is unclear

Welcome to the bifurcatoR wiki!

The idea is to keep notes here.

20220209

• Condense current dashboard down to 2 modules

  1. Detecting and describing bimodality
     • Drop down menu with multiple bimodality detection options (e.g. Hartigan's dip test, mclust bootstrapLRT, sig clust, etc)
     • Power to detect other features (e.g. if percentage of samples in each mode differs, if mode variances differ, other options)
  2. Comparing two distributions (can be uni modal or bi modal)
     • Have some pre-calcuated examples for efficiency (but still allow user to make selections on mode means and mixing percentages)
     • Prioritize variance tests (Levene's, Score, F, bootstrap, permutation, etc)
     • Test if the two distributions differ as a whole (KS test, anderson-darling, etc)
     • Test if mixing percentages differ significantly
     • Test if the distance between modes differs (delta-delta) 20220223

• Module 1 is coming along,we are going to drop tests/functionality to describe bimodality and only look at detecting bimodality. User will be able to select either mixture of Gaussians, or beta distribution. This way we avoid accidentally showing Mclust is best for everything.

• Module 2 will be focused on variance and distributional differences.