Closed philiptzou closed 2 years ago
The fix: since the geomean titer and the median fold are calculated separately - the geomean titer is from individual titers while the median fold is from individual folds from the division of each individual control titer by the individual experimental titer, but not the division of the geomean control titer by the experimental titer, there's not way to determine the cmp symbol consistent to geomean experimental titer without re-calculating the geomean experimental titer for the fold.
This introduced a quirk implementation which has been documented at https://github.com/hivdb/covid-drdb-payload/wiki/Method:-Use-potency-to-calcualte-fold-change-for-plasma-titer-(draft) as Quirk 1.
Quote from Bob:
The issue: for Planas21c vaccine AZD1222 the cmp symbols for titer and fold are inconsistent.
The issue is caused by the different calculation of aggregated titer and aggregated fold. The cmp symbol for aggregated titer is determined by compare geomean value with the largest titer threshold, plus the lower-than-threshold titers have been taken square-roots before they participating in geometric mean calculation. This method allows calling a "<" when a few number of experimental titers stay effective.
The cmp symbol for aggregated fold is determined by if all experimental titers are ineffective. This method doesn't call ">" when one or more of the experimental titers are effective.