Associations among hormones and anthropometric traits was tested using a a multivariate linear regression model . Multiple testing comparisons problem was taken into account using the False Discovery Rate (FDR) adjustment proposed by Benjamini and Hochberg.
Regressions with dichotomous outcomes were fitted using the glm function provided in the R package base. Regressions with count outcomes were fitted using the glm.nb function provided in the R package MASS.