Crosstabbing multiple response data

Thanks to everyone for this.  I was looking at the
same problem last night and just was going to write
a
posting to R-help when I saw this.  


--- Michael Wexler <wexler at yahoo.com> wrote:

Thanks to Charles, Gabor, and a private message

from

Frank E Harrell with some good ideas and help. 

This

crossprod approach was very clever, I would never
have thought of it.

Best, Michael


----- Original Message ----
From: Charles C. Berry <cberry at tajo.ucsd.edu>
To: Michael Wexler <wexler at yahoo.com>
Cc: r-help at stat.math.ethz.ch
Sent: Thursday, February 22, 2007 1:17:44 PM
Subject: Re: [R] Crosstabbing multiple response

data

res <- crossprod( as.matrix( ratings[ , -1] ) )
diag(res) <- ""
print(res, quote=F)

      att1 att2 att3
att1      2    1
att2 2         2
att3 1    2

res2 <- crossprod(as.matrix( ratings[ , -1])) *

100 / nrow( ratings )

res2[] <- paste( res2, "%", sep="" )
diag(res2) <- ""
print(res2, quote=F)

      att1 att2 att3
att1      50%  25%
att2 50%       50%
att3 25%  50%

Be sure to bone up on format and sprintf before
taking this into 
production.

On Thu, 22 Feb 2007, Michael Wexler wrote:

Using R version 2.4.1 (2006-12-18) on Windows, I

have a dataset which resembles this:

id    att1    att2    att3
1    1        1        0
2    1        0        0
3    0        1        1
4    1        1        1

ratings <- data.frame(id = c(1,2,3,4), att1 =

c(1,1,0,1), att2 = c(1,0,0,1), att3 = c(0,1,1,1))

I would like to get a cross tab of counts of

co-ocurrence, which might resemble this:

   att1    att2    att3
att1         2       1
att2    2            2
att3    1    2

with the hope of understanding, at least

pairwise,

what things "hang together".   (Yes, there are

much,

much better ways to do this statistically

including

clustering and binary corrected correlation, but

the

audience I am working with asked for this version
for a specific reason.)

(Later on, I would also like to convert to

percentages of the total unique pop, so the final
version of the table would be

   att1    att2    att3

att1         50%       25%

att2    50%            50%

att3    25%    50%


But I can do this in excel if I can get the

first

table out.)

I have tried the reshape library, but could not

get anything resembling this (both on its own, as
well as feeding in to table()).  (I have also

played

with transposing and using some comments from this
list from 2002 and 2004, but the questioners

appear

to assume more knowledge than I have in use of R;
the example in the posting guide was also more
complex than I was ready for, I'm afraid.)

Sample of some of my efforts:
library(reshape)
melt(ratings,id=c("id"))

ds1 <- melt(ratings,id=c("id"))
table(ds1$variable, ds1$variable) # returns only

rowcounts, 3 along diagonal

xtabs(formula = value ~ ds1$variable +

ds1$variable , data=ds1) # returns only a single

row

of collapsed counts, appears to not allow 1

variable

in multiple uses

I suspect I am close, so any nudges in the right

direction would be helpful.

Thanks much, Michael

PS: www.rseek.org is very impressive, I heartily

encourage its use.

    [[alternative HTML version deleted]]

protection around 
http://mail.yahoo.com