[R] Looping over groups

[Josh B]

When I try your suggestion, I get the output, e.g.,:


christina.df$gen: 309
NULL

Returned for each vector. That doesn't seem good (I associate "NULL" with something being empty), but perhaps I am misinterpreting it?

----- Original Message ----
From: Dylan Beaudette <dylan.beaudette at gmail.com>
To: Josh B <joshb41 at yahoo.com>
Cc: r-help at r-project.org
Sent: Wednesday, August 20, 2008 10:58:26 AM
Subject: Re: [R] Looping over groups

On Wed, Aug 20, 2008 at 7:48 AM, Josh B <joshb41 at yahoo.com> wrote:
> Hello,
>
> My R skills are somewhere between novice and intermediary, and I am hoping that some of you very helpful forum members, whom I've seen work your magic on other peoples' problems/questions, can help me here.
>
> I have a matrix with the following format:
>
> (i) individual plants comprising many different genotype groups (i.e., a plant is genotype 1 or genotype 2 or genotype 3, etc). The column for genotypes is called "gen", and the plants are members of genotype class 1 - 309, with no overlaps (i.e., you're either a genotype 1 or a genotype something else, but not both) and no missing values.
> (ii) Various trait measurements taken on the plants, with multiple replicates per genotype group
>
> I want to create a covariance matrix, separately for plants from each genotype group. I know how to use the command "cov"; my problem is that I have 309 different genotype groups and so I need to set up some sort of an automated loop to go through each genotype group and create a separate covariance matrix based on it.
>
> My question is, how do I make a loop to automatically go through and create these covariance matrices, i.e., a separate covariance matrix for plants from each genotype group?
>
> I am familiar with the "for" command, but I cannot get it to work. Here is my code:
>
> christina= read.table("christina.txt", sep= ",", na= "NA", header= TRUE)
> {for (i in 1:309)
> christina.i= subset(christina, gen == i)
> christina.i.clean= christina.i[,-1]
> christina.matrix.i= as.matrix(christina.i.clean)
> christina.cov.i= cov(christina.matrix.i, y= NULL, use= "complete.obs", method= c("pearson"))
> write.table(christina.cov.i, sep= ",", file= "covariances.csv", row.names= FALSE, col.names= FALSE, append= TRUE)}
>
>
> The problem occurs at my code snippet "gen == i". I want R to insert a number in place of "i", depending on what round of the loop it is on, but R insists that I am literally referring to a genotype class named "i". I have made sure that the column "gen" is numeric, but the same problem persists if I make the column a factor instead.
>
> Any help would be much appreciated, but help that includes sample code would be most useful. Thank you in advance!
>
> Sincerely,
> Josh
>

If you can make your data into a dataframe, you can use something like this:

# might work
christina.df <- as.data.frame(christina)

# this should work, if your ID var ('gen') is the first column in the data frame
by(christina.df, christina.df$gen, function(d) {
d.clean <- d[,-1]
cov(d.clean, y= NULL, use= "complete.obs", method= c("pearson")
} )

for more ideas, see ?by

Cheers,

Dylan