Compare Groups

setwd("C:/Users/mvx13/OneDrive - Texas State University/Papers/2025/00_NewPapers/Survey/PewSurvey/W152/Arka/V3")
library(readxl)
library(ggplot2)
library(dplyr)
library(compareGroups)
library(vtree)
library(DT)

df= read.csv("ATP W152 Data_cleanedFin_AC.csv")
df1= df[, c(1:13, 15)]
res1 <- compareGroups(DrCell ~ ., df1, ref = 1, max.ylev = 50,
                      max.xlev = 50,  chisq.test.perm=TRUE, chisq.test.B=10000)
res2= createTable(res1, show.ratio = TRUE)
res2

library(survey)

vars <- c("Area","Regn","Land","Age","Gender","Edu","Race","Born","Marr",
          "Politics","Income","Ideology","IntUse","DrCell","Wght")
df1 <- df[, vars]

# treat as categorical (edit if any should be numeric)
cat_vars <- setdiff(vars, "Wght")
df1[cat_vars] <- lapply(df1[cat_vars], function(x) factor(x, exclude = NULL))

# survey design (ids=~1 is OK if you only have weights)
des <- svydesign(ids = ~1, weights = ~Wght, data = df1)

# ---- helper: weighted % by group + survey-weighted chi-square p-value ----
wtab_cat <- function(var, by, design) {
  f <- as.formula(paste0("~", var, "+", by))
  tab <- svytable(f, design)                   # weighted counts
  colp <- prop.table(tab, margin = 2) * 100    # % within DrCell groups

  # design-based Rao-Scott test
  p <- tryCatch(svychisq(as.formula(paste0("~", var, "+", by)),
                        design, statistic = "F")$p.value,
                error = function(e) NA_real_)

  # return long data.frame
  out <- as.data.frame.matrix(round(colp, 2))
  out$Level <- rownames(out)
  rownames(out) <- NULL
  out$Variable <- var
  out$p_value <- p
  out[, c("Variable","Level", setdiff(names(out), c("Variable","Level")))]
}

# build table for all predictors (exclude DrCell and Wght)
preds <- setdiff(names(df1), c("DrCell","Wght"))
tbl_list <- lapply(preds, wtab_cat, by = "DrCell", design = des)
tbl_weighted <- do.call(rbind, tbl_list)

tbl_weighted

datatable(tbl_weighted)

# write.csv(tbl_weighted, "Table1_weighted.csv", row.names = FALSE)

single_var <- df %>%
  group_by(Age) %>%
  summarise(
    Count = n(),
    Weighted = sum(Wght),
    .groups = "drop"
  ) %>%
  mutate(
    Unweighted_Percent = 100 * Count / sum(Count),
    Weighted_Percent   = 100 * Weighted / sum(Weighted)
  )

single_var

two_var <- df %>%
  group_by(Regn, Land) %>%
  summarise(
    Count = n(),
    Weighted = sum(Wght),
    .groups = "drop"
  ) %>%
  group_by(Regn) %>%   # <-- cluster by first column
  mutate(
    Unweighted_Percent = 100 * Count / sum(Count),
    Weighted_Percent   = 100 * Weighted / sum(Weighted)
  ) %>%
  ungroup()
two_var