[R] Create Matrix from Loop of Vectors, Sort It and Pick Top-K

[Marc Schwartz]

on 06/19/2008 09:59 AM Gundala Viswanath wrote:
> Hi,
> 
> I have the following dataset (simplified for example).
> 
> __DATA__
> 300.35  200.25 104.30
> 22.00 31.12 89.99
> 444.50 22.10 43.00
> 22.10 200.55 66.77
> 
> Now from that I wish to do the following:
> 
> 1. Compute variance of each row
> 2. Pick top-2 row with highest variance
> 3. Store those selected rows for further processing
> 
> To achieve this, I tried to: a) read the table and compute
> variance  for each row, b) append variance with its original
> row in a vector, c) store a vector into multidimentional array (matrix),
> d) sort that array. But I am stuck at the step (b).
> 
> Can anybody suggest what's the best way to achieve
> my aim above?
> 
> This is the sample code I have so far (not working).
> 
> __BEGIN__
> 
> #data <- read.table("testdata.txt")
> 
> 
> # Is this a right way to initialize?
> all.arr = NULL
> 
> for (gi in 1:nofrow) {
>    gex <- as.vector(data.matrix(data[gi,],rownames.force=FALSE))
> 
>    #compute variance
>    gexvar <- var(gex)
> 
>    # join variance with its original vector
>    nvec <- c(gexvar,gex)
> 
>    # I'm stuck here.....This doesn't seem to work
>    all.arr <- data.frame(nvec)
> }
> 
> print(all.arr)
> __END__
> --

If your data is contained in a data frame 'DF':

 > DF
       V1     V2     V3
1 300.35 200.25 104.30
2  22.00  31.12  89.99
3 444.50  22.10  43.00
4  22.10 200.55  66.77


# Get row-wise variances and cbind() them to DF
 > DF.var <- cbind(DF, var = apply(DF, 1, var, na.rm = TRUE))

 > DF.var
       V1     V2     V3       var
1 300.35 200.25 104.30  9610.336
2  22.00  31.12  89.99  1361.915
3 444.50  22.10  43.00 56676.803
4  22.10 200.55  66.77  8622.817


# Sort DF by 'var' using order()
 > DF.var[order(DF.var$var, decreasing = TRUE), ]
       V1     V2     V3       var
3 444.50  22.10  43.00 56676.803
1 300.35 200.25 104.30  9610.336
4  22.10 200.55  66.77  8622.817
2  22.00  31.12  89.99  1361.915


To get the top 2, you can take a couple of approaches:

 > DF.var[order(DF.var$var, decreasing = TRUE)[1:2], ]
       V1     V2    V3       var
3 444.50  22.10  43.0 56676.803
1 300.35 200.25 104.3  9610.336


or

 > head(DF.var[order(DF.var$var, decreasing = TRUE), ], 2)
       V1     V2    V3       var
3 444.50  22.10  43.0 56676.803
1 300.35 200.25 104.3  9610.336


See ?cbind, ?apply, ?order and ?head for more information.

HTH,

Marc Schwartz