[R] unused factor levels in reshape

[Thomas Lumley]

On Wed, 22 Feb 2006, Daniel Farewell wrote:

> When reshaping a dataframe in which there are unused factor levels in 
> the id variable, I get the following error:
>
> Error in if (!all(really.constant)) warning(gettextf("some constant variables (%s) are really varying",  :
>        missing value where TRUE/FALSE needed

Yes, I think it's a bug. Thanks for pointing it out.

 	-thomas

>
> For example,
>
>> df <- data.frame(i = factor(rep(1:5, each = 2)), t = factor(rep(1:2, 5)), x = rep(rbinom(5, 1, 0.5), each = 2), y = rpois(10, 10))
>
>> subdf <- subset(df, i %in% 1:3)
>
> defines a dataframe, and a subframe with some unused factor levels (i = 4, 5). Then
>
>> reshape(df, v.names = "y", timevar = "t", idvar = "i", direction = "wide")
>  i x y.1 y.2
> 1 1 0  13   6
> 3 2 0  12   5
> 5 3 0  10   9
> 7 4 1   9  11
> 9 5 1  12   8
>
> works fine but
>
>> reshape(subdf, v.names = "y", timevar = "t", idvar = "i", direction = "wide")
> Error in if (!all(really.constant)) warning(gettextf("some constant variables (%s) are really varying",  :
>        missing value where TRUE/FALSE needed
>
> produces the error, which happens during the check to see if the variables assumed constant are constant. The problem is that reshape searches over all the levels of the id variable (i in this case) to see if the other variables (here x) are constant. But there is no x associated with i = 4, 5 in the smaller dataframe, so
>
>> tapply(subdf$x, subdf$i, function(x) length(unique(x)) == 1)
>   1    2    3    4    5
> TRUE TRUE TRUE   NA   NA
>
> produces some NAs. A slight change to the reshape code to work around this problem would be to use (the equivalent of)
>
>> tapply(subdf$x, subdf$i[, drop = TRUE], function(x) length(unique(x)) == 1)
>   1    2    3
> TRUE TRUE TRUE
>
> in the reshapeWide function within reshape, but perhaps there is a good reason not to do this?
>
> Daniel
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>

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