CODEWITHSUNDEEP
rmatrixdistance
user18441
user18441

Reputation: 663

how to extract intragroup and intergroup distances from a distance matrix? in R

I have this dataset:

values<-c(0.002,0.3,0.4,0.005,0.6,0.2,0.001,0.002,0.3,0.01)
codes<-c("A_1","A_2","A_3","B_1","B_2","B_3","B_4","C_1","C_2","C_3")
names(values)<-codes

In the codes, the letter indicates a group and the number a case within each group. Therefore I have three groups and 3 to 4 cases in each group (the actual dataset is much larger but this is a subset).

Then I calculate the distance matrix:

dist(values)->dist.m

Now I would like to convert the dist.m in a dataset with two columns: one containing the distances "inside" all groups (distance between A_1 and A_2, between B_2 and B_4, etc...), and another one containing the distances "between" groups (between A_1 and B_1, between C_1 and B_4, etc...)

Is there any easy way to do this in R?

Any help would be very appreciated.

thank you very much in advance.

Tina.

Upvotes: 3

Views: 2219

Answers (1)

IRTFM
IRTFM

Reputation: 263332

They may call them matrices but they are really not. There is however an as.matrix function that will let you get matrix indexing:

> as.matrix(dist.m)[grep("A", codes), grep("A", codes) ]
      A_1   A_2   A_3
A_1 0.000 0.298 0.398
A_2 0.298 0.000 0.100
A_3 0.398 0.100 0.000

So you can get the first part with pretty compact code:

> sapply(LETTERS[1:3], function(let) as.matrix(dist.m)[grep(let, codes), grep(let, codes) ]
+ )
$A
      A_1   A_2   A_3
A_1 0.000 0.298 0.398
A_2 0.298 0.000 0.100
A_3 0.398 0.100 0.000

$B
      B_1   B_2   B_3   B_4
B_1 0.000 0.595 0.195 0.004
B_2 0.595 0.000 0.400 0.599
B_3 0.195 0.400 0.000 0.199
B_4 0.004 0.599 0.199 0.000

$C
      C_1   C_2   C_3
C_1 0.000 0.298 0.008
C_2 0.298 0.000 0.290
C_3 0.008 0.290 0.000

Then use negative logical addressing to get the rest:

> sapply(LETTERS[1:3], function(let) as.matrix(dist.m)[grepl(let, codes), !grepl(let, codes) ]
+ )
$A
      B_1   B_2   B_3   B_4   C_1   C_2   C_3
A_1 0.003 0.598 0.198 0.001 0.000 0.298 0.008
A_2 0.295 0.300 0.100 0.299 0.298 0.000 0.290
A_3 0.395 0.200 0.200 0.399 0.398 0.100 0.390

$B
      A_1   A_2   A_3   C_1   C_2   C_3
B_1 0.003 0.295 0.395 0.003 0.295 0.005
B_2 0.598 0.300 0.200 0.598 0.300 0.590
B_3 0.198 0.100 0.200 0.198 0.100 0.190
B_4 0.001 0.299 0.399 0.001 0.299 0.009

$C
      A_1   A_2   A_3   B_1   B_2   B_3   B_4
C_1 0.000 0.298 0.398 0.003 0.598 0.198 0.001
C_2 0.298 0.000 0.100 0.295 0.300 0.100 0.299
C_3 0.008 0.290 0.390 0.005 0.590 0.190 0.009

I don't see a way of representing this as a two column data structure but you can use melt in pkg::reshape2 to get a three column structure:

> melt( as.matrix(dist.m)[grep("A", codes), grep("A", codes) ] )
  Var1 Var2 value
1  A_1  A_1 0.000
2  A_2  A_1 0.298
3  A_3  A_1 0.398
4  A_1  A_2 0.298
5  A_2  A_2 0.000
6  A_3  A_2 0.100
7  A_1  A_3 0.398
8  A_2  A_3 0.100
9  A_3  A_3 0.000

That would give you a rather long dataframe for display but it would be easy enough to put melt inside the function call.

Upvotes: 3

Related Questions