I have the following sample file:
"id";"PCA0";"PCA1";"PCA2"
1;6.142741644872954;1.2075898020608253;1.8946959360032403
2;-0.5329026419681557;-8.586870627925729;4.510113575138726
When I try to read it with:
d <- read.table("file.csv", sep=";", header=T)
id is a integer column, PCA0 a numeric an all subsequent columns are factors
class(d$iid)
[1] "integer"
class(d$PCA0)
[1] "numeric"
class(d$PCA1)
[1] "factor"
class(d$PCA2)
[1] "factor"
Why aren't the other columns numeric as well?
I know how to convert the columns, but I want my script to work without manually casting the types. Why doesn't R recognize the numeric columns?
This was a change make in R 3.1. There as been much discussion on the R-devel list about this. Basically if a number has too many digits, it's converted to a factor. This behavior is supposted be be reverted in 3.1.1 but no release date has been set as far as I know.
as @MrFlick says: too many digits.
you can force what you want by specifying colClasses argument:
read.table("test.csv",
sep=";",
header=TRUE,
colClasses=c("integer","numeric","numeric","numeric"))
if you really need as much precision as possible:
require(data.table)
d <- fread("test.csv")
Then modify to maximum precision stored:
d[,PCA0 := sprintf("%.15E",PCA0)]
d[,PCA1 := sprintf("%.15E",PCA1)]
d[,PCA2 := sprintf("%.15E",PCA2)]
gives:
> d
id PCA0 PCA1 PCA2
1: 1 6.142741644872954E+00 1.207589802060825E+00 1.8946959360032403
2: 2 -5.329026419681557E-01 -8.586870627925729E+00 4.510113575138726
note: fread should be smater + faster.
