Age Deconfounding
Partial Spearman correlations let feature labels reflect diagnosis, sex, APOE4, and comorbidities after removing age-related variance. Age deconfounding was necessary because raw age correlations dominated diagnosis correlations in the rep…
1 sources - 4 claims
Partial Spearman correlations let feature labels reflect diagnosis, sex, APOE4, and comorbidities after removing age-related variance. Age deconfounding was necessary because raw age correlations dominated diagnosis correlations in the reported analysis. Clinical annotation used partial Spearman correlations that controlled for age. Alive features were assigned clinical categories based on the strongest significant non-age age-partial correlation after FDR correction.