Given a data frame, how do I programmatically list only columns of a certain type (say, only 'factor', 'numeric', etc.)?
I created a function that receives data frames. I now need to do certain aggregations on the data frame but only for numeric columns and I do not know beforehand which columns are they. So, is there a way to programmatically examine the 'meta' information?
str() gives the structure of the object - but is there a way this information can be used to achieve the above?
For example,
sumNumerics
To select only factors variables :
>mydf[ , names(mydf)[ sapply(mydf,class)=="factor" ] ]
to select only logical variables :
>mydf[ , names(mydf)[ sapply(mydf,class)=="logical" ] ]
to select logical and factors variables :
>mydf[ , names(mydf)[ sapply(mydf,class) %in% c( "logical" , "factor" ) ] ]
You can also use sapply. Given the variable "df" is your data.frame, here are some hints:
sapply(df,is.numeric)
df[,sapply(df,is.numeric)]
colSums(df[,sapply(df,is.numeric)])
sapply(df,class)
classes = unique(sapply(df,class))
results.by.class = list()
for (aClass in classes) {
dfcols = df[,classes==aClass,drop=FALSE]
results.by.class[[aClass]]= switch(aClass,
"numeric" = colSums(dfcols),
"character" = apply(dfcols, 2, function(x) range(nchar(x))),
"factor" = apply(dfcols, 2, table))
}
results.by.class
[Edit: factor() -> unique(); I actually wanted to refer to the levels of the factor, but it is simpler to call "unique(.)" instead of "levels(factor(.))". The previos version could cycle several times though the same columns, what was not intended]
Great answers. Thanks all.
is.xxx() set of methods and sapply( , class) comparisions are very helpful.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package