Execute a formula for a function in R?
Any help with this would be really appreciated. I am using the Lumley poll package and am trying to simplify my code but hit it a bit.
The svymean function from the package in my code is called like this, the first argument is a formula indicating which variables I want, and the second argument is a dataset:
svymean(~hq_ehla, FraSvy, na.rm=TRUE)
I'm trying to create a function that will pull out the means (proportions) and standard errors for categorical variables, so I ran the following function:
stats <- function(repstat, num) {
estmean <- as.numeric(round(100 * repstat[num], digits=0))
estse <- round(100 * sqrt(attributes(repstat)$var[num,num]), digits=1)
return(list(mean=estmean, se=estse))
}
This works, so when I pull out the middle and one of my first category, for example, I use:
stats(svymean(~hq_ehla, FraSvy, na.rm=TRUE), 1)$mean
stats(svymean(~hq_ehla, FraSvy, na.rm=TRUE), 1)$se
What I would like to do is simplify this to something significantly shorter, where I may only have to write:
stats(FraSvy, "hq_ehla", 1)$mean
Or something like that. The problem is I can't figure out how to pass the formula to the function using the variable name.
source to share
You can use reformulate
to build a formula and call it svymean
in your function. Use ...
to pass na.rm
or other arguments tosvymean
stats <- function(terms, data, num, ...) {
.formula <- reformulate(terms)
repstat <- svymean(.formula, data, ...)
estmean <- as.numeric(round(100 * repstat[num], digits=0))
estse <- round(100 * sqrt(attributes(repstat)$var[num,num]), digits=1)
return(list(mean=estmean, se=estse))
}
stats(data = FraSvy, terms = "hq_ehla", 1, na.rm = TRUE)$mean
See this answer for more details on creating formula objects programmatically
Or you can pass a formula object inside a function.
stats2 <- function(formula, data, num, ...) {
repstat <- svymean(formula, data, ...)
estmean <- as.numeric(round(100 * repstat[num], digits=0))
estse <- round(100 * sqrt(attributes(repstat)$var[num,num]), digits=1)
return(list(mean=estmean, se=estse))
}
stats2(data = FraSvy, formula = ~hq_ehla, 1, na.rm = TRUE)$mean
source to share
Functions coef
and SE
can make your life easier.
# construct a function that takes the equation part of svymean as a string
# instead of as a formula. everything else gets passed in the same
# as seen by the `...`
fun <- function( var , ... ) svymean( reformulate( var ) , ... )
# test it out.
result <- fun( "hq_ehla" , FraSvy , na.rm = TRUE )
# print the results to the screen
result
# also your components
coef( result )
SE( result )
# and round it
round( 100 * coef( result ) )
round( 100 * SE( result ) )
source to share