R help - Jul 2005 - removing factor level represented by less than x rows

In a number of different situations I'm trying to
remove factor levels that are represented by less than
a certain number of rows, e.g. if I had the dataset aa
below and wanted to remove the species that are
represented in less than 2 rows:

data(iris)
aa <- iris[1:101,]

In this case, since I can see that the species
virginica only has one row, I can write:

table(aa$Species)
setosa versicolor  virginica 
        50         50          1 

aa[aa$Species != "virginica", ]

but:

aa[aa$Species == names(table(aa$Species)> 2),]

does not work.

This must be a fairly common task with a straight
forward solution that I can't see. Any ideas?

Best wishes,
Mikkel

Mikkel Grum wrote:
> In a number of different situations I'm trying to
> remove factor levels that are represented by less than
> a certain number of rows, e.g. if I had the dataset aa
> below and wanted to remove the species that are
> represented in less than 2 rows:
> 
> data(iris)
> aa <- iris[1:101,]
> 
> In this case, since I can see that the species
> virginica only has one row, I can write:
> 
> table(aa$Species)
> setosa versicolor  virginica 
>         50         50          1 
> 
> aa[aa$Species != "virginica", ]
> 
> but:
> 
> aa[aa$Species == names(table(aa$Species)> 2),]
> 
> does not work.
> 
If you take a look at "table(aa$Species) > 2" you'll see your
first
mistake. Namely, the names are all still present. Your second mistake is 
to use "==" to match two names. "==" does not work like
that. What you
want is "%in%" instead.

I think you want the following:

keep <- levels(aa$Species)[table(aa$Species) > 2]
aa <- aa[aa$Species %in% keep, ]

However, the level for "virginica" is still present in the Species 
variable. If you would like to drop this completely, then try

aa$Species <- aa$Species[drop = TRUE]

HTH,

--sundar

Mikkel Grum wrote:
> In a number of different situations I'm trying to
> remove factor levels that are represented by less than
> a certain number of rows, e.g. if I had the dataset aa
> below and wanted to remove the species that are
> represented in less than 2 rows:
> 
> data(iris)
> aa <- iris[1:101,]
> 
> In this case, since I can see that the species
> virginica only has one row, I can write:
> 
> table(aa$Species)
> setosa versicolor  virginica 
>         50         50          1 
> 
> aa[aa$Species != "virginica", ]
> 
> but:
> 
> aa[aa$Species == names(table(aa$Species)> 2),]
> 
> does not work.
> 
> This must be a fairly common task with a straight
> forward solution that I can't see. Any ideas?
> 
> Best wishes,
> Mikkel
library(Hmisc)
?combine.levels

This doesn't remove levels but combines infrequent ones though.

Frank

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University