I have a database with missing data. I need to impute data (I am using mice), then create new columns based on the original columns (using the imputed data). It is these new columns that I need to do my statistical analyses with.
Specifically, my participants filled in several questionnaires using a 7-point likert scale. Some didn't answer all the questions. I need to impute values, then 1- sum the values in the columns and have access to this new value for statistical analyses 2- depending on this sum, divide the participants into "mild, moderate, high" and use this for statistical analyses.
I have based what I am trying to do on this stackoverflow answer: Perform operation on each imputed dataset in R's MICE
Here is my code (using R):
# Create a sample bdd
bdd=data.frame(
gender=c("M","F","M", "M", "M", "F"),
choice=c(1,2,NA,1,1,1),
gardes=c(0,0,0,5,7,NA),
EE1=c(3,4,1,NA,3,0),
EE2=c(2,5,1,3,3,0),
EE3=c(3,NA,1,5,3,0),
EE4=c(3,6,1,2,3,0),
EE5=c(1,4,1,2,3,5),
EE6=c(3,1,1,3,3,4),
EE7=c(5,0,1,5,3,5),
EE8=c(2,6,1,1,3,3),
EE9=c(3,4,1,6,3,4)
)
# Create the additional variable - this will have missing values
bdd$EE <- bdd$EE1+bdd$EE2+bdd$EE3+bdd$EE4+bdd$EE5+bdd$EE6+bdd$EE7+bdd$EE8+bdd$EE9
# create ini to get access to meth and pred
ini <- mice(bdd, max = 0, print = FALSE)
# Change the method of imputation for EE, so that it always equals bdd$EE1+...+bdd$EE9
meth1 <- ini$meth
meth1["EE"] <- "~I(bdd$EE1+bdd$EE2+bdd$EE3+bdd$EE4+bdd$EE5+bdd$EE6+bdd$EE7+bdd$EE8+bdd$EE9)"
pred1 <- ini$pred
# change the predictor matrix so only bdd$EE1-9 predicts EE (necessary?)
pred1[ "EE", ] <- 0
pred1[ "EE", c("EE1", "EE2", "EE3", "EE4", "EE5", "EE6", "EE7", "EE8", "EE9")] <- 1
# change the predictor matrix so that EE isnt used to predict
pred1[ , "EE" ] <- 0
# Imputations
imput <- mice(bdd, seed=1, pred = pred1, meth = meth1, m=1, print = FALSE)
Please note that this does not work. Any other way to do this elegantly? TIA for any and all advice!!!
Edited to add: this is the error message I get when I try to run this code:
Warning messages:
1: In `[<-.data.frame`(`*tmp*`, , i, value = list(`1` = c(20L, 14L, :
replacement element 1 has 456 rows to replace 2 rows
2: In `[<-.data.frame`(`*tmp*`, , i, value = list(`1` = c(20L, 14L, :
replacement element 1 has 456 rows to replace 2 rows
3: In `[<-.data.frame`(`*tmp*`, , i, value = list(`1` = c(20L, 14L, :
replacement element 1 has 456 rows to replace 2 rows
4: In `[<-.data.frame`(`*tmp*`, , i, value = list(`1` = c(20L, 14L, :
replacement element 1 has 456 rows to replace 2 rows
5: In `[<-.data.frame`(`*tmp*`, , i, value = list(`1` = c(20L, 14L, :
replacement element 1 has 456 rows to replace 2 rows
Here is the bdd I created for this question:
gender choice gardes EE1 EE2 E3 EE4 EE5 EE6 E7 EE8 EE9
1 M 1 0 3 2 3 3 1 3 5 2 3
2 F 2 0 4 5 NA 6 4 1 0 6 4
3 M NA 0 1 1 1 1 1 1 1 1 1
4 M 1 5 NA 3 5 2 2 3 5 1 6
5 M 1 7 3 3 3 3 3 3 3 3 3
6 F 1 NA 0 0 0 0 5 4 5 3 4
Here's the code without the bug, after the correction that user20650 pointed out!