cluster_permute_test {CellaRepertorium} | R Documentation |
This tests a statistic for association between labels (for instance, cluster/clonal ID) and covariates (for instance, subject or treatment) by permuting the link between the two.
Each observation represents a cell.
statistic
is any function of labels
cluster_permute_test( ccdb, cell_covariate_keys, cell_label_key = ccdb$cluster_pk, cell_stratify_keys, statistic, n_perm, alternative = c("two.sided", "less", "greater"), sanity_check_strata = TRUE, ... )
ccdb |
|
cell_covariate_keys |
|
cell_label_key |
|
cell_stratify_keys |
optional |
statistic |
function of label (vector) and covariate (data.frame). Must return a scalar |
n_perm |
number of permutations to run |
alternative |
|
sanity_check_strata |
|
... |
passed to |
a list containing the observed value of the statistic, its expectation (under independence), a p-value, and the Monte Carlo standard error (of the expected value).
library(dplyr) # covariate should name one or more columns in `cell_tbl` cluster_idx = c(1, 1, 1, 2, 2, 3, 3) subject = c('A', 'A', 'B', 'B', 'B', 'C', 'C') contig_tbl = tibble(contig_pk = seq_along(cluster_idx), cluster_idx, subject) ccdb_test = ContigCellDB(contig_tbl = contig_tbl, contig_pk = 'contig_pk', cell_pk = c('contig_pk', 'subject', 'cluster_idx'), cluster_pk = 'cluster_idx') ccdb_test$cell_tbl cluster_permute_test(ccdb_test, 'subject', 'cluster_idx', statistic = purity, n_perm = 50)