test_that("No weights", { test_data <- readRDS(test_path("fixtures", "test_data.rds")) test_data$Y_M <- with(test_data, factor(findInterval(Y_C, quantile(Y_C, seq(0, 1, length = 5)), all.inside = TRUE))) set.seed(123) test_data$off <- runif(nrow(test_data)) test_data$clus <- sample(1:50, nrow(test_data), replace = TRUE) expect_no_condition({ fit0 <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), data = test_data) }) #M-estimation for mlogit expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), data = test_data, vcov = "HC0") }) expect_equal(coef(fit0), coef(fit)) expect_equal(vcov(fit0), vcov(fit)) fit_g <- mlogit::mlogit(Y_M ~ 0 | A * (X1 + X2 + X3 + X4 + X5), data = dfidx::dfidx(test_data, choice = "Y_M", shape = "wide")) ind <- unlist(split(seq_along(coef(fit0)), rep(seq_len(nlevels(test_data$Y_M) - 1), length(coef(fit0))/(nlevels(test_data$Y_M) - 1)))) expect_equal(unname(coef(fit0)), unname(coef(fit_g)[ind]), tolerance = 1e-7) expect_equal(unname(vcov(fit0)), unname(sandwich::sandwich(fit_g)[ind, ind]), tolerance = 1e-7) expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), cluster = ~clus, data = test_data) }) expect_equal(coef(fit0), coef(fit)) #Cluster-robust SEs expect_equal(unname(vcov(fit)), unname(sandwich::vcovCL(fit_g, cluster = test_data$clus)[ind, ind]), tolerance = 1e-7) #Offset expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5) + offset(off), data = test_data) }) expect_failure(expect_equal(coef(fit0), coef(fit))) }) test_that("Binary treatment", { test_data <- readRDS(test_path("fixtures", "test_data.rds")) test_data$Y_M <- with(test_data, factor(findInterval(Y_C, quantile(Y_C, seq(0, 1, length = 5)), all.inside = TRUE))) set.seed(123) test_data$off <- runif(nrow(test_data)) test_data$clus <- sample(1:50, nrow(test_data), replace = TRUE) expect_no_condition({ W <- weightit(A ~ X1 + X2 + X3 + X4 + X5 + X6 + X7 + X8 + X9, data = test_data, method = "glm", estimand = "ATE", include.obj = TRUE) }) expect_M_parts_okay(W) expect_no_condition({ fit0 <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), data = test_data, weightit = W) }) #M-estimation for mlogit expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), data = test_data, weightit = W, vcov = "asympt") }) expect_equal(coef(fit0), coef(fit)) expect_equal(vcov(fit0), vcov(fit)) expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), data = test_data, weightit = W, vcov = "HC0") }) mlogit_data <- dfidx::dfidx(transform(test_data, .weights = W$weights), choice = "Y_M", shape = "wide") fit_g <- mlogit::mlogit(Y_M ~ 0 | A * (X1 + X2 + X3 + X4 + X5), data = mlogit_data, weights = .weights, tol = 1e-12, ftol = 1e-12) ind <- unlist(split(seq_along(coef(fit)), rep(seq_len(nlevels(test_data$Y_M) - 1), length(coef(fit))/(nlevels(test_data$Y_M) - 1)))) expect_equal(unname(coef(fit)), unname(coef(fit_g)[ind]), tolerance = 1e-6) # expect_equal(unname(vcov(fit)), unname(sandwich::sandwich(fit_g)[ind, ind]), # tolerance = 1e-6) expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5), cluster = ~clus, data = test_data, weightit = W, vcov = "HC0") }) expect_equal(coef(fit0), coef(fit)) # #Cluster-robust SEs # expect_equal(unname(vcov(fit)), # unname(sandwich::vcovCL(fit_g, cluster = test_data$clus, cadjust = T)[ind, ind]), # tolerance = 1e-7) #Offset expect_no_condition({ fit <- multinom_weightit(Y_M ~ A * (X1 + X2 + X3 + X4 + X5) + offset(off), data = test_data) }) expect_failure(expect_equal(coef(fit0), coef(fit))) })