Error: calling arguments in user functions using quosures - r

Question

I am trying to create cross tables using weights::wtd.chi.sq.
The data:
data_in <- read_table2("Q50_1 Q50_2 Q38 Q90 pov gender wgt1 wgt2
never always Yes 2 High M 1.3 0.4
sometimes always No 2 Med F 0.4 0.2
always sometimes Yes 4 Low F 1.2 0.7
never never No 2 High M 0.5 0.7
always always No 4 High M 0.7 0.8
sometimes never Yes 3 Low F 0.56 0.3
sometimes never Yes 2 Med F 0.9 0.1
")
x_tab function that feeds into another function:
xtab_func <- function(dat, col, target, wgt){
col <- rlang::as_string(ensym(col))
target <- rlang::as_string(ensym(target))
wgt <- rlang::as_string(ensym(wgt))
wtd.chi.sq(dat[[target]],dat[[col]], weight = dat[[wgt]])
}
Running it gives:
xtab_func(data_in, 'Q50_1','pov','wgt1')
Chisq df p.value
7.3395092 4.0000000 0.1189981
Now I am looping through a vector of columns to repeat this for tabulation for each column. The error happens when I try to call target and wgt within the xtab function above. I've tried 3 different ways but none of them work.
crosstab <- function(dat, target, columns, wgt,target_name, school_type){
# browser()
target <- rlang::as_string(ensym(target))
print(target)
wgt <- rlang::as_string(ensym(wgt))
target_name <- enquo(target_name)
school_type <- enquo(school_type)
d <- list()
for (i in columns){
# OPTION 1
# x <- xtab_func(dat, i, !!target, !!wgt)
# OPTION 2
x <- xtab_func(dat, i, target, wgt)
# OPTION 3
# x <- xtab_func(dat, i, dat[[target]],dat[[wgt]])
x$i <- i
d[[i]] <- x
df <- do.call(rbind, d)
}
return(df)
}
When I run this I could see the chi values for the columns by pov...
cols <- data_in %>% select(starts_with("Q"))
cols <- names(cols)
crosstab(data_in,'pov',cols, 'wgt1', 'pov','trad')
But I get these errors:
Error: Only strings can be converted to symbols
OR
Error in model.frame.default(formula = weight ~ var1 + var2) :
invalid type (NULL) for variable 'var1'
Any idea how I call those variables? Thank you!

Answer 1

It's not clear to me at all why you are trying to use all the rlang stuff when you are just passing character values to your functions. This could be greatly simplified to
xtab_func <- function(dat, col, target, wgt){
weights::wtd.chi.sq(dat[[target]],dat[[col]], weight = dat[[wgt]])
}
and
crosstab <- function(dat, target, columns, wgt,target_name, school_type){
d <- list()
for (i in columns){
x <- as.data.frame(as.list(xtab_func(dat, i, target, wgt)))
x$i <- i
d[[i]] <- x
}
df <- do.call(rbind, d)
return(df)
}
Just use [[]] with character values to index into your data.

Answer 2

With the OP's default function xtab_func, we can modify the crosstab to
library(purrr)
library(dplyr)
crosstab <- function(dat, target, columns, wgt,target_name, school_type){
purrr::map_dfr(columns, ~ {
xtab_func(dat, !!.x, !!target, !!wgt)
})
}
-testing
crosstab(data_in,'pov', cols, 'wgt1', 'pov','trad')
# A tibble: 4 x 3
Chisq df p.value
<dbl> <dbl> <dbl>
1 7.34 4 0.119
2 6.02 4 0.198
3 1.47 2 0.480
4 4.83 4 0.306