tests/testthat/test-weighting.R

In Displayr/flipAnalysisOfVariance: Functions for computing ANOVA, MANOVA, and related methods

# context("weighting")
# preference <- factor(c(rep(c("Trump", "Clinton"), c(55, 45)), rep(c("Trump", "Clinton"), c(40, 60))))
# state <- factor(c(rep("Nevada", 100), rep("California", 100)))
# california <- 38800000 #Population
# nevada <- 2839000 #Population
# gross.weights <- c(rep(nevada / 100, 100), rep(california / 100, 100))
# dat <- data.frame(preference, state, gross.weights)
#
# test_that("sampling weights",{
#
#
#
# xtabs(~preference + state, data = dat)
#
# # Chi-square test
# summary(xtabs(~preference + state, data = dat))$p.value
#
# # z-test (performed using binary logit)
# summary(glm(preference ~ state, family = binomial, data = dat))$coef[2,4]
#
# # z-test using gross weights
#
# suppressWarnings(summary(glm(preference ~ state, family = quasibinomial, weights = gross.weights, data = dat))$coef[2,4])
#
# # z-test using weights with an average of 1 (i.e., which sum to the sample size)
# unit.weights <- gross.weights / mean(gross.weights) * 200
# summary(glm(preference ~ state, family = quasibinomial, weights = unit.weights, data = dat))$coef[2,4]
#
# # z-test using weights with an average of 1 (i.e., which sum to the sample size)
# unit.weights <- gross.weights / mean(gross.weights)
# summary(glm(preference ~ state, family = quasibinomial, weights = unit.weights, data = dat))$coef[2,4]
# xtabs(unit.weights ~ state)
#
# # Weight calibrated to the effective sample size
# ess <- flipData::EffectiveSampleSize(unit.weights)
# calibrated.weight <- unit.weights / sum(unit.weights) * ess
# summary(glm(preference ~ state, family = quasibinomial, weights = calibrated.weight, data = dat))$coef[2,4]
#
# # taylor series linearized weight
# calibrated.weight <- unit.weights / sum(unit.weights) * ess
# suppressWarnings(flipRegression::Regression(preference ~ state, type = "Binary Logit", weights = gross.weights, data = dat))
#
# suppressWarnings(flipAnalysisOfVariance::OneWayANOVA(dat$state, dat$preference, type = "Binary Logit", weights = gross.weights, correction = "None"))
#
# })