PCA downsampling not behaving as expected.

[KellyLBennett]

Using either the parameters cohorts and max_cohort_size , min_cohort_size with max_cohort_size or just cohort_size does not downsample per cohort but samples all individuals (see attached notebook for examples) and is therefore not behaving as expected.

Copy_of_Last_EA_PCA (2).ipynb

[KellyLBennett]

Here is a code example using cohorts and max_cohort_size.

sample_set=['1178-VO-UG-LAWNICZAK-VMF00025',
 '1246-VO-TZ-KABULA-VMF00185',
 '1246-VO-TZ-KABULA-VMF00197',
 '1270-VO-MULTI-PAMGEN-VMF00162',
 '1270-VO-MULTI-PAMGEN-VMF00218',
 '1274-VO-KE-KAMAU-VMF00246',
 '1324-VO-ET-GOLASSA-VMF00257',
 '1324-VO-ET-GOLASSA-VMF00275',
 'AG1000G-KE',
 'AG1000G-TZ',
 'AG1000G-UG',
 'fontaine-2015-rebuild']
sample_query='country==["Ethiopia","Tanzania","Uganda","Kenya"] and taxon=="arabiensis"'
region = "3L:15,000,000-41,000,000"
n_snps = 100_000

rm=['VBS58980-6781STDY12201453',
 'VBS59004-6781STDY12201477',
 'VBS59007-6781STDY12201480',
 'VBS59065-6781STDY12201540',
 'VBS59067-6781STDY12201542',
 'VBS59074-6781STDY12201549',
 'VBS59098-6781STDY12201573',
 'VBS59101-6781STDY12201576',
    'VBS94787-7466STDY14619753',
 'VBS94790-7466STDY14619756',
 'VBS94792-7466STDY14619758',
 'VBS94794-7466STDY14619760',
 'VBS94802-7466STDY14619768',
 'VBS94803-7466STDY14619769',
 'VBS94804-7466STDY14619770',
 'VBS94805-7466STDY14619771',
 'VBS94807-7466STDY14619773',
 'VBS94812-7466STDY14619778',
 'VBS94813-7466STDY14619779',
 'VBS94814-7466STDY14619780',
 'VBS94815-7466STDY14619781',
 'VBS94818-7466STDY14619784',
 'VBS94819-7466STDY14619785',
 'VBS94820-7466STDY14619786',
 'VBS94821-7466STDY14619787',
 'VBS94822-7466STDY14619788',
 'VBS94823-7466STDY14619789',
 'VBS94825-7466STDY14619791',
 'VBS94827-7466STDY14619793',
 'VBS94828-7466STDY14619794',
 'VBS94830-7466STDY14619796',
    'VBS80612-7281STDY13749335',
 'VBS80631-7281STDY13749350',
 'VBS80637-7281STDY13749354',
 'VBS80638-7281STDY13749355',
 'VBS80665-7281STDY13749377',
 'VBS80718-7281STDY13749432',
 'VBS80728-7281STDY13749442',
 'VBS80738-7281STDY13749452',
 'VBS80786-7281STDY13749502',
 'VBS80787-7281STDY13749503',
 'VBS80860-7281STDY13749565',
 'VBS83477-7466STDY14207192',
 'VBS83481-7466STDY14207196',
 'VBS95022-7466STDY14620152',
 'VBS95039-7466STDY14620169',
 'VBS95040-7466STDY14620170',
 'SAMN01760608']

pca_East_AfriAll_outlier_rm_df,evr =ag3.pca(region=region,n_snps=n_snps,
                     n_components=20,thin_offset=0,
                     sample_sets=sample_set,
                     cohorts='location',
                     sample_query=sample_query,
                     sample_query_options=None,
                     sample_indices=None,
                     site_mask="arab",
                     site_class=None,
                     min_minor_ac=2,
                     max_missing_an=0,
                     cohort_size=None,
                     min_cohort_size=None,
                     max_cohort_size=50,
                     exclude_samples=rm,
                     fit_exclude_samples=None,
                     random_seed=42,
                     inline_array=True,
                     chunks="native")

Only a total of 50 individuals from across all cohorts are selected.