anova_stats: Effect size statistics for anova

Description

Returns the (partial) eta-squared, (partial) omega-squared, epsilon-squared statistic or Cohen's F for all terms in an anovas. anova_stats() returns a tidy summary, including all these statistics and power for each term.

Usage

anova_stats(model, digits = 3)
cohens_f(model)
epsilon_sq(model, ci.lvl = NULL, n = 1000, method = c("dist",
  "quantile"))
eta_sq(model, partial = FALSE, ci.lvl = NULL, n = 1000,
  method = c("dist", "quantile"))
omega_sq(model, partial = FALSE, ci.lvl = NULL, n = 1000,
  method = c("dist", "quantile"))

Arguments

model

A fitted anova-model of class aov or anova. Other models are coerced to anova.

digits

Number of decimal points in the returned data frame.

ci.lvl

Scalar between 0 and 1. If not NULL, returns a data frame with effect sizes including lower and upper confidence intervals.

Number of bootstraps to be generated.

method

Character vector, indicating if confidence intervals should be based on bootstrap standard error, multiplied by the value of the quantile function of the t-distribution (default), or on sample quantiles of the bootstrapped values. See 'Details' in boot_ci(). May be abbreviated.

partial

Logical, if TRUE, the partial eta-squared is returned.

Value

A data frame with the term name(s) and effect size statistics; if ci.lvl is not NULL, a data frame including lower and upper confidence intervals is returned. For anova_stats(), a tidy data frame with all statistics is returned (excluding confidence intervals).

Details

For eta_sq() (with partial = FALSE), due to non-symmetry, confidence intervals are based on bootstrap-methods. In this case, n indicates the number of bootstrap samples to be drawn to compute the confidence intervals. Confidence intervals for partial omega-squared and epsilon-squared is also based on bootstrapping.

Since bootstrapped confidence intervals are based on the bootstrap standard error (i.e. mean(x) +/- qt(.975, df = length(x) - 1) * sd(x)), bounds of the confidence interval may be negative. Use method = "quantile" to make sure that the confidence intervals are always positive.

References

Levine TR, Hullett CR (2002): Eta Squared, Partial Eta Squared, and Misreporting of Effect Size in Communication Research (pdf)

Tippey K, Longnecker MT (2016): An Ad Hoc Method for Computing Pseudo-Effect Size for Mixed Model. (pdf)

Examples

Run this code

# NOT RUN {
# load sample data
data(efc)

# fit linear model
fit <- aov(
  c12hour ~ as.factor(e42dep) + as.factor(c172code) + c160age,
  data = efc
)

eta_sq(fit)
omega_sq(fit)
eta_sq(fit, partial = TRUE)
eta_sq(fit, partial = TRUE, ci.lvl = .8)

anova_stats(car::Anova(fit, type = 2))

# }

Run the code above in your browser using DataLab

Get 50% off unlimited learning