Separately, I’m still running Mathematica version of fitting linear regression to logit-transformed frequencies to demarcate our >0.05 per day growth in frequency threshold
Could we write a small script to scan forecast-ncov results for this threshold? It really should be clade-in-question vs “other” however.
Should be able to convert relative growth advantage to per-day growth rate (site: delta in models)
Multiply Pango lineage frequency by relative fitness will give mean fitness
How would this work for regions for which we have no countries included, e.g. South America/Africa? Using this flaggin strategy would increase clade bias towards high income/sequencing countries. See #64
Context
Requested by @trvrb in today's forecasting-automation meeting: