How might I write a function that takes a string vector and returns a function with arguments having the names of the string vector? In addition, I'd like to use my string of arguments to make a list or data.frame named with the input columns inside that function. My desired application is to be able to pass this to a predict method for estimating some points. If this function already exists, let me know. Otherwise, I'm curious how I might write it. Below I include R pseudo-code illustrating what I am trying to do and hopefully showing where I getting stuck conceptually.
make_fitting_function <- function(mod) {
x <- xterms(mod)
function(!! x) {
predict(mod, newdata = list(!! x))
}
}
Calling this function returns a function that can be called on separate vector arguments. For example:
f <- make_fitting_function(lm(mpg ~ wt, data = mtcars))
f(wt = c(1, 2, 3, 4, 5))
The result would be:
1 2 3 4 5
31.94 26.60 21.25 15.91 10.56
However, this would also work for many more x variables (e.g.):
f <- make_fitting_function(lm(mpg ~ wt + am + carb, data = mtcars))
f(wt = c(1, 2, 3, 4, 5), am = rep(1, 5), carb = seq(2, 10, by = 2))
Ideally, this function should be able to be used by integrate in the case of a single variable.
You actually don't need to name the parameters; instead just make the data frame to pass in to predict using all of them.
make_fitting_function <- function(model) {
function(...) {
predict(model, newdata=data.frame(...))
}
}
f <- make_fitting_function(lm(mpg ~ wt + am + carb, data = mtcars))
f(wt = c(1, 2, 3, 4, 5), am = rep(1, 5), carb = seq(2, 10, by = 2))
## 1 2 3 4 5
## 30.589985 24.637415 18.684846 12.732277 6.779708