Search code examples
rdataframescaledata-cleaningstandardized

How to avoid scaling dummy variables in dataframe in r?

I want to standardise all my variables before applying machine learning methods. However, to my understanding, dummy variables should never be standardised. After entering the following code, r standardized all my variables, even the ones which are binary. How can I avoid this happening?

#standardize all non-categorical variables to have mean zero and a standard deviation of one

df_standardized <- df %>% mutate(across(where(is.numeric), scale))

I checked my data types are they are "int", not numeric. Thank you in advance for your help.

Solution

  • scale returns a matrix, we can convert the matrix to vector by either as.numeric or as.vector. In addition, use inherits for only modifying the numeric columns

    library(dplyr)
out <- df %>% 
   mutate(across(where(~ inherits(.x, "numeric")),
         ~ as.numeric(scale(.x))))

    data

    data(iris)
df <- iris
df$intCol <- 1L