R - user defined function to replace missing values - r

Question

I am trying to create a user-defined function to replace missing values in each variable using specific probabilities.
I can get the probabilities to print, but the second part of the code does not seem to work and all the missing values still remain.
I do not get any particular error message so puzzled why it is not working.
My data:
library(dplyr)
library(glue)
structure(list(id = c("395891", "373742", "316241", "282072",
"341331", "251761", "154591", "125051", "095361", "141822", "281411",
"31571", "165191", "03212", "08091", "26172", "135561", "164331",
"344511", "37352"), ph201_01 = c(1L, NA, 1L, 1L, NA, 1L, 1L,
NA, NA, NA, 1L, NA, NA, 1L, NA, 1L, NA, NA, 1L, NA), ph201_02 = c(NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_), ph201_03 = c(NA,
NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 1L, NA, NA, NA,
NA, NA, NA), ph201_04 = c(NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_,
NA_integer_, NA_integer_)), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -20L))
My code:
# Create user-defined function
create_mock_vars = function(var) {
# get prevalence
prev = round(sum(mydata[[var]], na.rm=TRUE)/nrow(mydata), 2)
print(glue("Prevalence of {var} is {prev}."))
mydata[[var]][is.na(mydata[[var]])] = sample(0:1, size=sum(is.na(mydata[[var]])), replace=TRUE, prob=c(prev, 1-prev))
return(mydata)
}
# Get list of variable names I want to impute
myvarnames = names(mydata[,-1])
# Apply my function
sapply(myvarnames, create_mock_vars)
glimpse(mydata)

Answer 1

You can create the user-defined function as :
create_mock_vars = function(x) {
prev <- sum(x, na.rm=TRUE)/length(x)
x[is.na(x)] <- sample(0:1, size = sum(is.na(x)), replace=TRUE,
prob= c(prev, 1-prev))
return(x)
}
and use lapply to apply it to each myvarnames columns
mydata[myvarnames] <- lapply(mydata[myvarnames], create_mock_vars)

Answer 2

Not entirely sure what you mean so here are two Base R solutions:
df1 <- data.frame(lapply(df,
function(x) {
if (is.numeric(x) & sum(is.na(x)) == length(x)) {
NA_integer_
} else if (is.numeric(x)) {
ifelse(is.na(x), sum(x, na.rm = TRUE) / length(x), x)
} else{
ifelse(is.na(x), na.omit(x)[cumsum(!is.na(x))], x)
}
}))
df2 <- data.frame(lapply(df,
function(x) {
if (is.numeric(x)) {
ifelse(sum(is.na(x)) == length(x), 1,
ifelse(sum(is.na(x)) != length(x),
sum(x, na.rm = TRUE) / length(x), x))
} else{
ifelse(is.na(x), na.omit(x)[cumsum(!is.na(x))], x)
}
}))
Data:
df <- structure(
list(
id = c("395891", "373742", "316241", "282072",
"341331", "251761"),
ph201_01 = c(1L, NA, 1L, 1L, NA, 1L),
ph201_02 = c(
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_
),
ph201_03 = c(
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_
),
ph201_04 = c(
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_,
NA_integer_
)
),
class = c("tbl_df",
"tbl", "data.frame"),
row.names = c(NA, -6L)
)

Answer 3

I was able to solve this but without creating a user-defined function. Still, it would be great if someone could help me figure out what I did wrong for future reference?
myvarnames = names(mydata[,-1])
for (i in myvarnames) {
prev = round(sum(mydata[[i]], na.rm=TRUE)/nrow(mydata), 2) # prevalence of deficit
mydata[[i]][is.na(mydata[[i]])] = sample(0:1, size=sum(is.na(mydata[[i]])), replace=TRUE, prob=c(prev, 1-prev))
}
glimpse(mydata)