Choose variables based on name (simple regular expression)

Question

I would like to incorporate variable names that imply what I should do with them. I imagine a dataframe "survey".

library(Rlab) # Needed for rbern() function.
survey <- data.frame(cbind(  
id = seq(1:10),  
likert_this = sample(seq(1:7),10, replace=T),  
likert_that = sample(seq(1:7), 10, replace=T),  
dim_bern_varx = rbern(10, 0.6),  
disc_1 = sample(letters[1:5],10,replace=T)))

Now I would like to do certain things with all variables that contain likert, other things with variables that contain bern etc.

How can this be done in R?

Answer 1

You can use grep() with colnames():

survey[,grep("bern", colnames(survey))]

Answer 2

If you have a series of names you like to grab you can also use match. perhaps you often need variables "pulse", "exercise", "height", "weight" and "age", but they sometimes show up in different places or with other added variables. You can save the vector of common names then match them against the dataframe and have a new df of just your standard columns in the order you want.

basenames <- c("pulse", "exercise", "height", "weight", "age")
get.columns <- match(basenames, names(dataframe))
new.df <- dataframe[,get.columns]

Answer 3

The "operators" package allows some Perl-like syntax:

library(operators)

survey[, colnames(survey) %~% "bern"]

or

subset(survey, select = colnames(survey) %~% "bern")