Weighted ANOVA across multiple outcomes — survey

Fits survey::svyglm() for each outcome against a categorical grouping variable and runs survey::regTermTest(). Optionally returns pairwise contrasts via emmeans.

Usage

survey_anova(
  svy_df,
  outcomes,
  group_var,
  wt_var,
  pairwise = FALSE,
  adjust = "bonferroni"
)

Arguments

svy_df: A data frame containing the survey data.
outcomes: Character vector of (continuous) outcome variables.
group_var: A string naming the categorical grouping variable. Will be coerced to factor.
wt_var: A string naming the survey weight column.
pairwise: If TRUE, also return pairwise contrasts from emmeans::emmeans().
adjust: Multiple-comparison adjustment for pairwise contrasts. See ?emmeans::summary.emmGrid.

Value

A list with element overall (one row per outcome: outcome, group_var, statistic, df_num, df_den, p_value) and, if pairwise = TRUE, element pairwise (one row per contrast: outcome, contrast, estimate, se, df, t_ratio, p_value).

Details

Requires the survey package. Pairwise contrasts also require emmeans.

Examples

if (FALSE) { # \dontrun{
survey_anova(nptrends_y6_clean,
             outcomes = "LLLostAmt",
             group_var = "SizeStrata",
             wt_var = "weight_year6plus",
             pairwise = TRUE)
} # }