Variable Selection for Mixed Data Clustering: Application in Human Population Genomics