Repository navigation
WLS regression workflow #51
Description
We want to support the types of analyses that were possible in Enrich2, especially the weighted least squares regression and replicate combination using the random effects model.
This is likely to require some new plugins (like a random effects model plugin), but that's something we can use down the line to (for example) combine scores from individual barcodes into a single variant-level score.
The example workflow can target the existing Enrich2 example dataset, which already is downsampled to include a limited number of variants.
Activity
nickzoic commentedon Apr 21, 2026
Random Effects Model code is in Enrich2 at https://github.com/FowlerLab/Enrich2/blob/main/enrich2/random_effects.py and documentation of the algorithm is in "Mixed Models: Theory and Applications with R", Eugene Demidenko, 2 ed, pp 233 - 239.
The Enrich2 code implements this as an operation on an entire dataframe but for CountESS we'll implement this as a python function operating on a single row and use
DuckdbParallelTransformPlugin.transformto run it over all rows.nickzoic commentedon Apr 29, 2026
Support for this is working now in v0.0.19 and getting the correct answers!