[英]Storing simulation results in R
我想估計Mantel-Haenszel差異項功能(DIF)賠率和HMDDIF指數。 我在下面編寫了函數。 在我看來,存儲結果時我犯了一個錯誤。 您能看看這個並給我反饋嗎? 這是示例數據:
# generate dataset
r <- 1000
c <- 16
test <- matrix(rbinom(r*c,1,0.5),r,c)
# create sum scores for each student using first 15 columns
test <- cbind(test, apply(test[,1:15],1,sum))
colnames(test) <- c("v1","v2","v3","v4","v5","v6","v7","v8","v9","v10","v11","v12","v13","v14","v15","group","score")
test <- as.data.frame(test)
前15列是學生對項目/問題的正確/錯誤答案。 組成員資格列是第16列。 學生的“得分”變量是最后一列(第17列)的項目得分總和。 在我從Wikipedia( https://en.wikipedia.org/wiki/Differential_item_functioning )獲得的圖片中,可以找到該公式。
對於每個分數類別,我想估計這張圖中的最后兩個公式。 行是10個學生,列是六個項目/問題。 同樣,第16列是組成員身份(1-焦點,0引用)這是我的功能代碼。
library(dplyr)
# this function first starts with the first item and loop k scores from 1-15. Then move to the second item.
# data should only contain the items, grouping variable, and person score.
Mantel.Haenszel <- function (data) {
# browser() #runs with debug
for (item in 1:15) { #item loop not grouping/scoring
item.incorrect <- data[,item] == 0
item.correct <- data[,item] == 1
Results <- c()
for (k in 1:15) { # for k scores
Ak <- nrow(filter(data, score == k, group == 0, item.correct)) # freq of ref group & correct
Bk <- nrow(filter(data, score == k, group == 0, item.incorrect)) # freq of ref group & incorrect
Ck <- nrow(filter(data, score == k, group == 1, item.correct)) # freq of foc group & correct
Dk <- nrow(filter(data, score == k, group == 1, item.incorrect)) # freq of foc group & incorrect
nrk <- nrow(filter(data, score == k, group == 0)) #sample size for ref
nfk <- nrow(filter(data, score == k, group == 1)) #sample size for focal
if (Bk == 0 | Ck == 0) {
next
}
nominator <-sum((Ak*Dk)/(nrk + nfk))
denominator <-sum((Bk*Ck)/(nrk + nfk))
odds.ratio <- nominator/denominator
if (odds.ratio == 0) {
next
}
MH.D.DIF <- (-2.35)*log(odds.ratio) #index
# save the output
out <- list("Odds Ratio" = odds.ratio, "MH Diff" = MH.D.DIF)
results <- rbind(Results, out)
return(results)
} # close score loop
} # close item loop
} #close function
這就是我得到的
# test funnction
Mantel.Haenszel(test)
> Mantel.Haenszel(test)
Odds Ratio MH Diff
out 0.2678571 3.095659
我想要得到的是
> Mantel.Haenszel(test)
Odds Ratio MH Diff
out 0.2678571 3.095659
## ##
.. ..
(15 rows here for 15 score categories in the dataset)
您不應該為item
和k
每種組合期望結果,最大輸出行數為225,除非有break
任何實例? 如果是這樣,我認為您只需要更改一些小事情。 首先,在函數開始時只聲明一次Results
。 然后,確保您正在rbind
並返回Results
或results, but not both. Then, move your
返回results, but not both. Then, move your
results, but not both. Then, move your
返回值results, but not both. Then, move your
到實際功能級別,而不是循環。 在下面的示例中,我還包括了當前item
和k
進行演示:
Mantel.Haenszel <- function (data) {
# browser() #runs with debug
Results <- c()
for (item in 1:15) {
#item loop not grouping/scoring
item.incorrect <- data[, item] == 0
item.correct <- data[, item] == 1
for (k in 1:15) {
# for k scores
Ak <-
nrow(filter(data, score == k, group == 0, item.correct)) # freq of ref group & correct
Bk <-
nrow(filter(data, score == k, group == 0, item.incorrect)) # freq of ref group & incorrect
Ck <-
nrow(filter(data, score == k, group == 1, item.correct)) # freq of foc group & correct
Dk <-
nrow(filter(data, score == k, group == 1, item.incorrect)) # freq of foc group & incorrect
nrk <-
nrow(filter(data, score == k, group == 0)) #sample size for ref
nfk <-
nrow(filter(data, score == k, group == 1)) #sample size for focal
if (Bk == 0 | Ck == 0) {
next
}
nominator <- sum((Ak * Dk) / (nrk + nfk))
denominator <- sum((Bk * Ck) / (nrk + nfk))
odds.ratio <- nominator / denominator
if (odds.ratio == 0) {
next
}
MH.D.DIF <- (-2.35) * log(odds.ratio) #index
# save the output
out <-
list(
item = item,
k = k,
"Odds Ratio" = odds.ratio,
"MH Diff" = MH.D.DIF
)
Results <- rbind(Results, out)
} # close score loop
} # close item loop
return(Results)
} #close function
test.output <- Mantel.Haenszel(test)
給出類似的輸出:
> head(test.output, 20)
item k Odds Ratio MH Diff
out 1 3 2 -1.628896
out 1 4 4.666667 -3.620046
out 1 5 0.757085 0.6539573
out 1 6 0.5823986 1.27041
out 1 7 0.9893293 0.02521097
out 1 8 1.078934 -0.1785381
out 1 9 1.006237 -0.01461145
out 1 10 1.497976 -0.9496695
out 1 11 1.435897 -0.8502066
out 1 12 1.5 -0.952843
out 2 3 0.8333333 0.4284557
out 2 4 2.424242 -2.08097
out 2 5 1.368664 -0.7375117
out 2 6 1.222222 -0.4715761
out 2 7 0.6288871 1.089938
out 2 8 1.219512 -0.4663597
out 2 9 1 0
out 2 10 2.307692 -1.965183
out 2 11 0.6666667 0.952843
out 2 12 0.375 2.304949
那是您要找的東西嗎?
聲明:本站的技術帖子網頁,遵循CC BY-SA 4.0協議,如果您需要轉載,請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.