I want to create a 2X2 table for Chisq test from multiple levels categorical dataset-CodePudding

I have a dataset of race and outcome either (Y,N) I want to tabulate a 2X2 table to run a chisq test for each race.

  Asian     584   24
  Black    1721   56
  Hispanic 2400   90
  White    8164  289

Once I create a table 2X2 so the first row will be Asian and second row will be non-Asian (counted from values of total - asianNo) and (total- asian yes) as the second column of that row. Then I can run a chisq test easily on each race once I repeat that process for all races. Is there an easier way to run a Chisq test for each race in my table above?

CodePudding user response：

Here is one option with map2, where the first row is an individual race and the second row are the others, then I name each list according to the specific race.

library(tidyverse)

pull(df, V1) %>%
  map2(
    .,
    replicate(nrow(df), df, simplify = FALSE),
    .f = function(x, y)
      y %>%
      filter(V1 != x) %>%
      summarise(across(-V1, sum)) %>%
      bind_rows(filter(y, V1 == x) %>% dplyr::select(-V1), .)
  ) %>%
  set_names(., pull(df, V1))

Output

$Asian
     V2  V3
1   584  24
2 12285 435

$Black
     V2  V3
1  1721  56
2 11148 403

$Hispanic
     V2  V3
1  2400  90
2 10469 369

$White
    V2  V3
1 8164 289
2 4705 170

Data

df <- structure(list(V1 = c("Asian", "Black", "Hispanic", "White"), 
    V2 = c(584L, 1721L, 2400L, 8164L), V3 = c(24L, 56L, 90L, 
    289L)), class = "data.frame", row.names = c(NA, -4L))

CodePudding user response：

Here is another approach. First set up the master table:

tbl <- as.matrix(df[, -1])
Sums <- matrix(colSums(tbl), nrow(tbl), 2, byrow=TRUE)
Tbl <- cbind(tbl, Sums-tbl)
row.names(Tbl) <- df[, 1]
Tbl
#           Yes  No   Yes  No
# Asian     584  24 12285 435
# Black    1721  56 11148 403
# Hispanic 2400  90 10469 369
# White    8164 289  4705 170

Now a function to create 2x2 tables from a row in Tbl:

ChiSqTable <- function(row) {
    matrix(Tbl[row, ], 2, 2, byrow=TRUE, dimnames=list(Race=c(df[row, 1],
         paste("Not", df[row, 1])), Question=c("Yes", "No")))
}

Finally create Chi Square tables and run the test:

Tables <- lapply(seq(nrow(Tbl)), ChiSqTable)
names(Tables) <- df[, 1]
ChiSqStats <- lapply(Tables, chisq.test)
names(ChiSqStats) <- df[, 1]

Tables[[1]]   # or Tables[["Asian"]]
#            Question
# Race          Yes  No
#   Asian       584  24
#   Not Asian 12285 435
ChiSqStats[[1]]
# 
#   Pearson's Chi-squared test with Yates' continuity correction
# 
# data:  X[[i]]
# X-squared = 0.33997, df = 1, p-value = 0.5598

Access the remaining tables, statistical results by specifying the number or Race. All of the results of the Chi Square Test are saved, e.g.

ChiSqStats[[1]]$expected
#            Question
# Race               Yes        No
#   Asian       587.0612  20.93878
#   Not Asian 12281.9388 438.06122
ChiSqStats[[1]]$residuals
#            Question
# Race                Yes         No
#   Asian     -0.12634367  0.6689899
#   Not Asian  0.02762242 -0.1462607