Detect contingency table type for a pair of categorical variables

For a given pair of columns, this function determines whether the cross-tabulation yields a 2x2 or a larger (n x m) contingency table, after removing rows with missing values in either column (pairwise deletion). This drives the dispatch between the phi coefficient and Cramer's V in effect_size.

Usage

detect_type(x, y)

Arguments

x

A factor, character, or logical vector.

y

A factor, character, or logical vector of the same length as x.

Value

A character string: "2x2" if both variables have exactly two levels observed in the pairwise-complete data, or "RxC" otherwise.

Details

Level counts are computed on the pairwise-complete subset, not on the full vector. This means that a variable with three levels in the full data may be classified as having two levels for a specific pair if one level is entirely missing when the other variable is observed.

See also

effect_size, compute_assoc

Examples

x <- c("a", "b", "a", "b", NA)
y <- c("yes", "no", "yes", "yes", "no")
detect_type(x, y)  # "2x2"
#> [1] "2x2"

z <- c("low", "mid", "high", "low", "mid")
detect_type(x[-5], z[-5])  # "RxC"
#> [1] "RxC"