Estimate models with groups of variables

Estimate models with groups of variables.R

# Create sample data
set.seed(1)
data <- data.frame(y = rnorm(10),
                   a = rnorm(10),
                   b = rnorm(10),
                   c = rnorm(10),
                   d = rnorm(10))

# Function takes a string
# for outcome, character vector of variable
# names to always keep, character vector of
# variable names to add, and data
quick_lm <- function(outcome, keepers, new, data){
  keepers <- paste(keepers, collapse = " + ")
  new <- paste(new, collapse =  " + ")
  rhs <- paste(keepers, new, sep = " + ")
  model <- paste0(outcome, " ~ ", rhs)
  fit <- lm(as.formula(model), data = data)
}

# Arbitrarily assign some variables to inputs
outcome <- "y"
keepers <- c("a","b")

# Fit a model, check summary
m1 <- quick_lm(outcome = outcome, keepers = keepers, new = "c", data = data)
summary(m1)

# Same model, shorter version
m1 <- quick_lm(outcome, keepers, "c", data)

# Try some other variable
m2 <- quick_lm(outcome = outcome, keepers = keepers, new = "d", data = data)
summary(m2)

# Same model, shorter version
m2 <- quick_lm(outcome, keepers, "d", data)

# Try a vector of variables
m3 <- quick_lm(outcome = outcome, keepers = keepers, new = c("c","d"), data = data)

# Same model, shorter version
m3 <- quick_lm(outcome, keepers, c("c","d"), data)

# If you really want a sparse function, just hard-code in
# the variables you want for your outcome and keepers:

my_outcome <- "y"
my_keepers <- c("a","b")
my_data <- data

# Hard-coded variable functions
quick_lm <- function(outcome = my_outcome, keepers = my_keepers, new = NULL, data = my_data){
  keepers <- paste(keepers, collapse = " + ")
  new <- paste(new, collapse =  " + ")
  rhs <- paste(keepers, new, sep = " + ")
  model <- paste0(outcome, " ~ ", rhs)
  fit <- lm(as.formula(model), data = data)
}

# Same as m1, m2, m3, but less code due to
# less flexible function
m4 <- quick_lm(new = "c")
m5 <- quick_lm(new = "d")
m6 <- quick_lm(new = c("c","d"))