R help - Feb 2005 - Row median of Date class variables in a data frame

I am trying to calculate the median of each row of a
data frame where the data frame consist of
columns of class Date.

Below are my test data and best attempt at using apply.
I didn't see a solution via Google or the Baron search
site.

I'd be grateful for any suggestions or solutions.
I'm using R 2.0.0 on Mac OS X.

Thank you,

Stephen Weigand


### Test data

date1 <- c(1000, 2000, 3000,4000)
date2 <- date1 + 100
date3 <- date2 + 100

class(date1) <- class(date2) <- class(date3) <- "Date"

test <- data.frame(date1, date2, date3)

print(test)

### create a function for apply()
medDate <- function(x){
   obj <- unclass(unlist(x))
   med <- median(obj, na.rm = TRUE)
   med
   class(med) <- "Date"
   med
}

medDate(test$date1) # works
medDate(test[1,])   # works

apply(test, 1, medDate) # gives error: 'need numeric data'
apply(test, 2, medDate) # gives error: 'need numeric data'

Stephen D. Weigand <weigand.stephen <at> charter.net> writes:

: 
: I am trying to calculate the median of each row of a
: data frame where the data frame consist of
: columns of class Date.
: 
: Below are my test data and best attempt at using apply.
: I didn't see a solution via Google or the Baron search
: site.
: 
: I'd be grateful for any suggestions or solutions.
: I'm using R 2.0.0 on Mac OS X.
: 
: Thank you,
: 
: Stephen Weigand
: 
: ### Test data
: 
: date1 <- c(1000, 2000, 3000,4000)
: date2 <- date1 + 100
: date3 <- date2 + 100
: 
: class(date1) <- class(date2) <- class(date3) <- "Date"
: 
: test <- data.frame(date1, date2, date3)
: 
: print(test)
: 
: ### create a function for apply()
: medDate <- function(x){
:    obj <- unclass(unlist(x))
:    med <- median(obj, na.rm = TRUE)
:    med
:    class(med) <- "Date"
:    med
: }
: 
: medDate(test$date1) # works
: medDate(test[1,])   # works
: 
: apply(test, 1, medDate) # gives error: 'need numeric data'
: apply(test, 2, medDate) # gives error: 'need numeric data'
: 

Try this:

library(zoo)
as.Date(apply(data.matrix(test), 1, median)) # ignore warning

The only reason you need zoo is that it supplies as.Date methods
which convert numerics and integers to Date.

Note that the argument to apply() is a matrix, not a data frame, so your
object has been coerced to a matrix.  That is spelled out on the help page 
for apply(), as it is frequent misuse.

You would be better off using a matrix of class Date in the first place if 
you want row operations.


On Wed, 23 Feb 2005, Stephen D. Weigand wrote:
> I am trying to calculate the median of each row of a
> data frame where the data frame consist of
> columns of class Date.
>
> Below are my test data and best attempt at using apply.
> I didn't see a solution via Google or the Baron search
> site.
>
> I'd be grateful for any suggestions or solutions.
> I'm using R 2.0.0 on Mac OS X.
>
> Thank you,
>
> Stephen Weigand
>
>
> ### Test data
>
> date1 <- c(1000, 2000, 3000,4000)
> date2 <- date1 + 100
> date3 <- date2 + 100
>
> class(date1) <- class(date2) <- class(date3) <- "Date"
>
> test <- data.frame(date1, date2, date3)
>
> print(test)
>
> ### create a function for apply()
> medDate <- function(x){
>  obj <- unclass(unlist(x))
>  med <- median(obj, na.rm = TRUE)
>  med
>  class(med) <- "Date"
>  med
> }
>
> medDate(test$date1) # works
> medDate(test[1,])   # works
>
> apply(test, 1, medDate) # gives error: 'need numeric data'
> apply(test, 2, medDate) # gives error: 'need numeric data'
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595