more on the daisy function

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

more on the daisy function

Adrian Dusa-2

Dear R-helpers,

First of all, a happy new year to everyone!

I succesfully used the daisy function (from package cluster) to find which two
rows from a dataframe differ by only one value, and I now want to come up with
a simpler way to find _which_ value makes the difference between any such
pair of two rows.
Consider a very small example (the actual data counts thousands of rows):

   input <- matrix(letters[c(1,2,1,2,2,3,2,1,1,2,2,2)], ncol=3)

   > input
     X1 X2 X3
   1  a  b  a
   2  b  c  b
   3  a  b  b
   4  b  a  b

I am interested by the rows which differ by one value only; I easily do that
with:

   library(cluster)
   distance <- daisy(as.data.frame(input))*ncol(input)

   > distance
   Dissimilarities :
     1 2 3
   2 3
   3 1 2
   4 3 1 2

   Metric :  mixed ;  Types = N, N, N
   Number of objects : 4


The first and the third rows differ only with respect to variable V3, and the
second and the fourth rows differ only with respect to variable V2.


Now I want to replace the different values by an "x"; currently my code is:

   distance <- as.matrix(distance)
   distance[!upper.tri(distance)] <- NA
   to.be.compared <- as.matrix(which(distance == 1, arr.ind=T))
   logical.result <- t(apply(to.be.compared, 1,
                       function(idx) {input[idx[1], ] == input[idx[2], ]}))
   result <- t(sapply(1:nrow(to.be.compared),
                       function(idx) {input[to.be.compared[idx, 1], ]}))
   result[!logical.result] <- "x"

   > as.data.frame(result)
     V1 V2 V3
   1  a  b  x
   2  b  x  b

I wonder if the daisy function could be persuaded to output a similar object
as the dissimilarities one; it would be fantastic to also get something like:

   First.difference.found:
     1 2 3
   2 1
   3 3 1
   4 1 2 1

Here, 3 means the third variable (V3) that the first and third rows differ on.
I could try to do that myself, but I don't know where to find the Fortran
code daisy uses.

Thanks for any hint,
Adrian

--
Adrian DUSA
Romanian Social Data Archive
1, Schitu Magureanu Bd
050025 Bucharest sector 5
Romania
Tel./Fax: +40 21 3126618 \
          +40 21 3120210 / int.101

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html