How do I include GLM in a loop in R

Question

My GLM models works fine, but in order to speed things up I want to include the variable names in a loop and make R calculate the GLM for each variable. So I tried this:

varlist <- c("age", "doc")

for (i in 1:length(varlist))
{

glmmodel <- glm(formula = Kommunikation ~ varlist[i], family = binomial, data = analysis_data)

univars[i,1] <- names(coef(glmmodel))[2]
univars[i,2] <- exp(confint.default(glmmodel)[2,1])
univars[i,3] <- exp(glmmodel$coefficients[2])
univars[i,4] <- exp(confint.default(glmmodel)[2,2])

}

Unfortunately, this results in the error:

Error in model.frame.default(formula = Kommunikation ~ varlist[[i]], data = analysis_data,  : 
  variable lengths differ (found for 'varlist[[i]]')

As the GLM model works when I substitute varlist[i] with the respective variable names age and doc , I would guess the issue stems from how the characters are read by R when substituting the variable name? (I have 26 variables but posting two only for convenience.)

Answer 1

The problem is that you're trying to drop a string into your formula in your call to glm() . Here's one possible solution in which you pass the string to a call to as.formula and then use that formula in the model instead.

df <- data.frame(y = rbinom(10, 1, 0.5), x1 = rnorm(10), x2 = rnorm(10))

varlist <- c("x1", "x2")

univars <- data.frame()  # create an empty data frame so the rest of your code works

for (i in seq_along(varlist))
{

mod <- as.formula(sprintf("y ~ %s", varlist[i]))

glmmodel <- glm(formula = mod, family = binomial, data = df)

univars[i,1] <- names(coef(glmmodel))[2]
univars[i,2] <- exp(confint.default(glmmodel)[2,1])
univars[i,3] <- exp(glmmodel$coefficients[2])
univars[i,4] <- exp(confint.default(glmmodel)[2,2])

}

Result:

> univars
  V1        V2        V3        V4
1 x1 0.4728192 3.1185658 20.569074
2 x2 0.1665581 0.7241709  3.148592

Also, I would be inclined to do this with lapply instead of a for loop, making rows that you can then bind. But this works.