Data frames with mixed data types

Question

I have been using R for a little while, but I am still struggling with factors and data frames. Here's my question.

I am trying to pre-allocate a data frame composed of several columns of different types, as follows:

cb <- data.frame(S=character(1000), I=numeric(1000), A=as.Date(rep(0,1000), origin = "1900-01-01"), SD=as.POSIXct(rep(0,1000), origin = "1900-01-01 00:00:00"), CC=numeric(1000), stringsAsFactors=FALSE)

which gets met the data frame types that I want (output of str(cb)):

'data.frame':   1000 obs. of  5 variables:
 $ S : chr  "" "" "" "" ...
 $ I : num  0 0 0 0 0 0 0 0 0 0 ...
 $ A : Date, format: "1900-01-01" "1900-01-01" "1900-01-01" "1900-01-01" ...
 $ SD: POSIXct, format: "1900-01-01" "1900-01-01" "1900-01-01" "1900-01-01" ...
 $ CC: num  0 0 0 0 0 0 0 0 0 0 ...

When I assign the first item in the data frame, CC and I become characters:

cb[1, ] <- c("ABCD", 4, "2005-12-12", "2008-04-03 20:30", 3)

output of str(cb):

'data.frame':   1000 obs. of  5 variables:
 $ S : chr  "ABCD" "" "" "" ...
 $ I : chr  "4" "0" "0" "0" ...
 $ A : Date, format: "2005-12-12" "1900-01-01" "1900-01-01" "1900-01-01" ...
 $ SD: POSIXct, format: "2008-04-03 20:30:00" "1900-01-01 00:00:00" "1900-01-01 00:00:00" "1900-01-01 00:00:00" ...
 $ CC: chr  "3" "0" "0" "0" ...

which makes it rather unusable for my purposes.

When I omit stringsAsFactors=FALSE in the data.frame definition, I (obviously) get a different error message (having set warn to 2):

Error in `[<-.factor`(`*tmp*`, iseq, value = "ABCD") : 
  (converted from warning) invalid factor level, NAs generated

which I understand but I am not sure how to overcome either.

What am I doing wrong? How can I make sure to keep the numeric type for columns I and SD? Thanks so much for your help.

Cheers

B

Answer 1

You can't mix types in a vector, so your vector is being coerced to character.

R> c("ABCD", 4, "2005-12-12", "2008-04-03 20:30", 3)
[1] "ABCD"             "4"               
[3] "2005-12-12"       "2008-04-03 20:30"
[5] "3"

[<-.data.frame then coerces the numeric columns of your data.frame to character, so the column will be one type; though I find it a bit inconsistent that it doesn't also convert the Date/POSIXt fields to character as well...

You can mix types in a list. This replacement works because data.frames are lists underneath.

cb[1, ] <- list("ABCD", 4, "2005-12-12", "2008-04-03 20:30", 3)

When you look back at your code later, it might make more sense to replace one row of your data.frame with a 1-row data.frame:

cb[1, ] <- data.frame("ABCD", 4, "2005-12-12", "2008-04-03 20:30", 3,
                      stringsAsFactors=FALSE)