R devel - Mar 2009 - 'sep' argument in reshape()

I wonder if the 'sep' argument in reshape() is being ignored
unintentionally:

## From example(reshape)
 df <- data.frame(id=rep(1:4,rep(2,4)),
                  visit=I(rep(c("Before","After"),4)),
                  x=rnorm(4), y=runif(4))

reshape(df, timevar="visit", idvar="id",
direction="wide", sep = "_")

  id x.Before y.Before x.After y.After
1  1    0.773    0.293  -0.021   0.658
3  2   -0.518    0.351  -0.623   0.946
5  3    0.773    0.293  -0.021   0.658
7  4   -0.518    0.351  -0.623   0.946

Is this more of the intended result when 'sep = "_"'?

  id x_Before y_Before x_After y_After
1  1    0.773    0.293  -0.021   0.658
3  2   -0.518    0.351  -0.623   0.946
5  3    0.773    0.293  -0.021   0.658
7  4   -0.518    0.351  -0.623   0.946

Thanks,

Stephen
-- 
Rochester, Minn. USA

>>>>> "SW" == Stephen Weigand <weigand.stephen at
gmail.com>
>>>>>     on Tue, 31 Mar 2009 18:33:05 -0500 writes:
    SW> I wonder if the 'sep' argument in reshape() is being ignored
    SW> unintentionally:

No.  It is used much differently than you *assume* it's used.

As always,   ?reshape   contains the answer.


    SW> ## From example(reshape)
    SW> df <- data.frame(id=rep(1:4,rep(2,4)),
    SW> visit=I(rep(c("Before","After"),4)),
    SW> x=rnorm(4), y=runif(4))

    SW> reshape(df, timevar="visit", idvar="id",
direction="wide", sep = "_")

    SW> id x.Before y.Before x.After y.After
    SW> 1  1    0.773    0.293  -0.021   0.658
    SW> 3  2   -0.518    0.351  -0.623   0.946
    SW> 5  3    0.773    0.293  -0.021   0.658
    SW> 7  4   -0.518    0.351  -0.623   0.946

    SW> Is this more of the intended result when 'sep =
"_"'?

    SW> id x_Before y_Before x_After y_After
    SW> 1  1    0.773    0.293  -0.021   0.658
    SW> 3  2   -0.518    0.351  -0.623   0.946
    SW> 5  3    0.773    0.293  -0.021   0.658
    SW> 7  4   -0.518    0.351  -0.623   0.946

no it is not.

I tend to agree that I would have preferred a different argument
name than 'sep' for the current 'sep',
and then a *further* argument 'sep' with the functionality that
you'd like would be straightforward.

Martin Maechler, ETH Zurich

On Tue, 31 Mar 2009, Stephen Weigand wrote:
> I wonder if the 'sep' argument in reshape() is being ignored
> unintentionally:
>
> ## From example(reshape)
> df <- data.frame(id=rep(1:4,rep(2,4)),
>                  visit=I(rep(c("Before","After"),4)),
>                  x=rnorm(4), y=runif(4))
>
> reshape(df, timevar="visit", idvar="id",
direction="wide", sep = "_")
>
>  id x.Before y.Before x.After y.After
> 1  1    0.773    0.293  -0.021   0.658
> 3  2   -0.518    0.351  -0.623   0.946
> 5  3    0.773    0.293  -0.021   0.658
> 7  4   -0.518    0.351  -0.623   0.946
>
> Is this more of the intended result when 'sep = "_"'?
No. sep= is designed for going the other way.  If you have wide-format data with
variable names x.Before y.Before x.After y.After, using sep="." will
let reshape() work out that the long-format variable names are x and y and the
conditions to be put in the time variable are Before and After.

    -thomas


Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle