R help - Dec 2010 - melt causes errors when characters and values are used

Hello,

I am finding that the melt function from the reshape library causes
errors when applied to a data.frame that contains numeric and character
columns.  For example,

melt(id.vars="ID",data.frame(ID=1:3,date=c("a","b","c"),value=c(1,4,5)))
  ID variable value
1  1     date     a
2  2     date     b
3  3     date     c
4  1    value  <NA>
5  2    value  <NA>
6  3    value  <NA>
Warning message:
In `[<-.factor`(`*tmp*`, ri, value = c(1, 4, 5)) :
  invalid factor level, NAs generated

It would be useful in this situation that the numerical column got
converted to a character column in this situation.  Any ways round this?

In actual fact I have got a situation where it is more like this

ID	Date_1	Value_1	Date_2	Value_2 ...

and I would like to convert it to a data.frame of ID, Date & Value but I
thought the above would be an appropriate middle step.

Thanks

Dan
-- 
**************************************************************
Daniel Brewer, Ph.D.

Institute of Cancer Research
Molecular Carcinogenesis
Email: daniel.brewer at icr.ac.uk
**************************************************************

The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company
Limited by Guarantee, Registered in England under Company No. 534147 with its
Registered Office at 123 Old Brompton Road, London SW7 3RP.

This e-mail message is confidential and for use by the a...{{dropped:2}}

Hi:


On Fri, Dec 10, 2010 at 6:47 AM, Daniel Brewer
<daniel.brewer@icr.ac.uk>wrote:
> Hello,
>
> I am finding that the melt function from the reshape library causes
> errors when applied to a data.frame that contains numeric and character
> columns.  For example,
>
In this case, thankfully so - it's warning that you're trying to mix
numeric
and factor variables.
>
>
melt(id.vars="ID",data.frame(ID=1:3,date=c("a","b","c"),value=c(1,4,5)))
>  ID variable value
> 1  1     date     a
> 2  2     date     b
> 3  3     date     c
> 4  1    value  <NA>
> 5  2    value  <NA>
> 6  3    value  <NA>
> Warning message:
> In `[<-.factor`(`*tmp*`, ri, value = c(1, 4, 5)) :
>  invalid factor level, NAs generated
>
> It would be useful in this situation that the numerical column got
> converted to a character column in this situation.  Any ways round this?
>
For this question, convert value to character before melting it.
>
> In actual fact I have got a situation where it is more like this
>
> ID      Date_1  Value_1 Date_2  Value_2 ...
>
> and I would like to convert it to a data.frame of ID, Date & Value but
I
> thought the above would be an appropriate middle step.
>
This is a more effective way, taken from a post on the ggplot2 list,
courtesy of Kohske Takahashi:

# Toy data:
df <- data.frame(id = as.character(101:110),
                  date1 = seq(from = as.Date('2006-01-10'),
                                to = as.Date('2006-01-19'), by =
'day'),
                   value1 = rpois(10, 5),
                  date2 = seq(from = as.Date('2006-02-10'),
                                to = as.Date('2006-02-19'), by =
'day'),
                   value2 = rpois(10, 10),
                   stringsAsFactors = FALSE)

dfm <- data.frame(
            melt(df, id = 'id', measure = c(grep('^date',
names(df)))),
            myval = melt(df, id = 'id',
                         measure = c(grep('^val', names(df))))$value
                 )
names(dfm)[3] <- 'date'

The first melt() call creates the three variables that are normally returned
from melt(): id, variable and value. The second, internal melt() call
creates a stacked column for the values without creating a new
'variable'.
One can repeat the game with multiple sets of variables.

Of course, there is an alternative way to reshape such data with the
reshape() function...


HTH,
Dennis


Thanks
>
> Dan
> --
> **************************************************************
> Daniel Brewer, Ph.D.
>
> Institute of Cancer Research
> Molecular Carcinogenesis
> Email: daniel.brewer@icr.ac.uk
> **************************************************************
>
> The Institute of Cancer Research: Royal Cancer Hospital, a charitable
> Company Limited by Guarantee, Registered in England under Company No.
534147
> with its Registered Office at 123 Old Brompton Road, London SW7 3RP.
>
> This e-mail message is confidential and for use by the...{{dropped:13}}