Convenience helpers that run chi_square and return just the
requested effect size as a numeric value (named by group for grouped
data): phi() (sqrt(chi-squared / N); in a 2x2 table signed like
SPSS's Phi, negative when the first row goes with the second column;
bounded by 1 only in 2x2 tables), cramers_v() (normalized to 0-1
for any table size), and
goodman_gamma() for two ordinal variables. As in SPSS, Phi and
Cramer's V are computed for any table size.
For the full test output (chi-square statistic, p-value, all effect
sizes), call chi_square directly.
Usage
phi(data, ..., weights = NULL)
cramers_v(data, ..., weights = NULL)
goodman_gamma(data, ..., weights = NULL)Arguments
- data
Your survey data (a data frame or tibble)
- ...
Exactly two categorical variables, as in
chi_square()- weights
Optional survey weights. Give a column name (unquoted or as a string), an expression such as
sampling_weight * 2, or a numeric vector with one weight per row.
Examples
data(survey_data)
phi(survey_data, gender, region)
#> [1] 0.01288807
cramers_v(survey_data, education, region)
#> [1] 0.01964626
goodman_gamma(survey_data, education, life_satisfaction)
#> [1] 0.292086