Skip to contents

Test theme co-occurrence patterns with chi-square tests of independence

Source: R/14_correlations.R
test_theme_cooccurrence.Rd

For each pair of themes, tests whether co-occurrence is significantly different from expected by chance.

Usage

test_theme_cooccurrence(
  data,
  theme_set,
  min_expected = 5,
  min_theme_entries = 5L,
  min_observed_both = 1L
)

Arguments

data

Tibble with theme_membership_* columns

theme_set

ThemeSet object

min_expected

Minimum expected cell count for chi-square (default 5)

min_theme_entries

Integer; themes with fewer than this many positive entries are excluded. Default 5L, matching the correlation matrix + theme-group test default.

min_observed_both

Integer; polish. Pairs whose observed co-occurrence is below this count are skipped (Fisher tests on zero-co-occurrence pairs are uninterpretable). Default 1L.

Value

Tibble with co-occurrence test results

Details

applies the same min_theme_entries filter that prepare_correlation_data and compare_theme_groups use, so the three statistical layers report counts over a consistent theme cohort. An earlier version admitted every theme regardless of frequency, which produced thousands of degenerate Fisher tests on rare themes (a large run found 99.1% of Fisher pairs had observed_both = 0).