volker (version 3.2.0)

effect_counts_one_grouped: Output test statistics and effect size for contingency tables

Description

Chi squared is calculated using stats::chisq.test. If any cell contains less than 5 observations, the exact-parameter is set.

Usage

effect_counts_one_grouped(data, col, cross, clean = TRUE, ...)

Value

A volker list with two volker tibbles. The first tibble contains npmi values for each combinations:

• n Number the combination occurs.

• p_x Marginal share of the first category.

• p_y Marginal share of the second category.

• p_xy Share of the combination based on all combinations.

• ratio The ratio of p_xy to (p_x * p_y).

• pmi Pointwise Mutual information, derived from the ratio.

• npmi Normalized Pointwise Mutual Information, derived from the pmi.

The second tibble contains effect sizes based on the cross table:

• Cramer's V: Effect size measuring the association between the two variables.

• n: Number of cases the calculation is based on.

• Chi-squared: Chi-Squared test statistic. If expected values are below 5 in at least one cell, an exact Fisher test is conducted.

• df: Degrees of freedo of the chi-squared test. Empty for the exact Fisher test.

• p: p-value of the chi-squared test.

• stars: Significance stars based on p-value (*, **, ***).

Arguments

data

A tibble.

col

The column holding factor values.

cross

The column holding groups to compare.

clean

Prepare data by data_clean.

...

Placeholder to allow calling the method with unused parameters from effect_counts.

Details

Phi is derived from the Chi squared value by sqrt(fit$statistic / n). Cramer's V is derived by sqrt(phi / (min(dim(contingency)[1], dim(contingency)[2]) - 1)). Cramer's V is set to 1.0 for diagonal contingency matrices, indicating perfect association.

Examples

library(volker)
data <- volker::chatgpt

effect_counts_one_grouped(data, adopter, sd_gender)