R: Apply a list with glm-function-CodePudding

I'm having trouble trying to put variables, which are in list, into a glm-function. My dataframe has lots of variables, so it would be too much effort to type the independent variables one by one into glm. Lets say my dataframe is

df <- data.frame(
  id = c(1,2,3,4,5),
  `sitting position` = c("A","B","A","A","B"),
  `variable one` = c("left", "left", "right", "right", "left"),
  `variable two` = c(50, 30, 45, 80, 57),
  `variable three` = c("m","w","w","m","m"),
  check.names = FALSE)

and my list of columns which I want to use in a glm-function looks like this

columns <- dput(colnames(df))[-c(1:2)]

columns 
[1] "variable one"   "variable two"   "variable three"

Now I want to put this list directly into a glm- function, something like

glm(`sitting position` ~ columns, data = df, familiy = binomial).

instead of

glm(`sitting position` ~ `variable one`   `variable two`   `variable three`, data = df, family = binomial())

I am aware that I can't work just by adding the list, but I also can't find a solution to fix this problem.

CodePudding user response：

Maybe we can use reformulate. reformulate will create formulas from character vectors. We can feed the output of reformulate into the formula argument to the glm function.

I included a preliminary step to replace your column names with a cleaner and less buggy alternative with janitor::clean_names.

library(janitor)

df<-df %>% clean_names
columns<-c('variable_one', 'variable_two', 'variable_three')

And then the actual solution:

glm(formula=reformulate(termlabels = columns, response='sitting_position'), data=df)

See how reformulate works:

reformulate(termlabels = columns, response='sitting_position')

sitting_position ~ variable_one   variable_two   variable_three