How to read csv file in R where some values contain the percent symbol (%)

Question

Is there a clean/automatic way to convert CSV values formatted with as percents (with trailing % symbol) in R?

Here is some example dat

Answer 1

There is no "percentage" type in R. So you need to do some post-processing:

DF <- read.table(text="actual,simulated,percent error
2.1496,8.6066,-300%
0.9170,8.0266,-775%
7.9406,0.2152,97%
4.9637,3.5237,29%", sep=",", header=TRUE)

DF[,3] <- as.numeric(gsub("%", "",DF[,3]))/100

#  actual simulated percent.error
#1 2.1496    8.6066         -3.00
#2 0.9170    8.0266         -7.75
#3 7.9406    0.2152          0.97
#4 4.9637    3.5237          0.29

Answer 2

Tidyverse has multiple ways of solving such issues. You can use the parse_number() specification which will strip a number off any symbols, text etc.:

sample_data = "actual,simulated,percent error\n 2.1496,8.6066,-300%\n 0.9170,8.0266,-775%\n7.9406,0.2152,97%\n4.9637,3.5237,29%"
DF <- read_csv(sample_data,col_types = cols(`percent error`= col_number()))

# A tibble: 4 x 3
# actual   simulated `percent error`
# <chr>        <dbl>           <dbl>
# 1 2.1496     8.61           -300  
# 2 + 0.9170     8.03           -775 
# 3 + 7.9406     0.215            97.0
# 4 + 4.9637     3.52             29.0 

Answer 3

With data.table you can achieve it as

a <- fread("file.csv")[,`percent error` := as.numeric(sub('%', '', `percent error`))/100]

Answer 4

This is the same as Roland's solution except using the stringr package. When working with strings I'd recommend it though as the interface is more intuitive.

library(stringr)
d <- str_replace(junk$percent.error, pattern="%", "")
junk$percent.error <- as.numeric(d)/100