[R] weighted mean and by() with two index

[William Dunlap]

Note that that output of by() is a matrix, but with some extra
attributes added to it.

Since you didn't supply any data I made up some that might
resemble yours.

  > set.seed(1)
  > re<-list(meta.sales.lkm=data.frame(pc=runif(40), sales=rpois(40,3),
size=sample(c("small","medium","large"),size=40,replace=TRUE),
yr=sample(1994:1998,size=40,replace=TRUE)))

I ran your by() call

   > tmp<-by(re$meta.sales.lkm[, c("pc", "sales")],
               re$meta.sales.lkm[, c("size", "yr")],
               function(x) weighted.mean(x[,1], x[,2]))

and looked at it with dput and saw that all the usual
components of a matrix are in it

   > dput(tmp)
   structure(c(0.86969084572047, 0.687022846657783, 0.40032217082464,
   0.125555095961317, 0.529131081343318, 0.64538708513137,
0.613078526553831,
   0.663822646145351, 0.48206098045921, 0.333916208640273,
0.513083046752339,
   NA, 0.457996427547187, 0.30292882991489, NA), .Dim = c(3L, 5L
   ), .Dimnames = structure(list(size = c("large", "medium", "small"
   ), yr = c("1994", "1995", "1996", "1997", "1998")), .Names =
c("size",
   "yr")), call = by.data.frame(data = re$meta.sales.lkm[, c("pc",
       "sales")], INDICES = re$meta.sales.lkm[, c("size", "yr")],
       FUN = function(x) weighted.mean(x[, 1], x[, 2])), class = "by")

It is just the print method for 'by' objects that makes it look
different.
Since there is no special 'by' method for '[' you can use tmp[,] to view
the matrix part of it

   > tmp[,]
           yr
   size          1994      1995      1996      1997      1998
     large  0.8696908 0.1255551 0.6130785 0.3339162 0.4579964
     medium 0.6870228 0.5291311 0.6638226 0.5130830 0.3029288
     small  0.4003222 0.6453871 0.4820610        NA        NA

If there were a [.by then you might have to manually remove the
"call" attribute and change the class to "matrix".

Bill Dunlap
TIBCO Software Inc - Spotfire Division
wdunlap tibco.com 

---------------------------------------
R] weighted mean and by() with two index

Dong H. Oh r.arecibo at gmail.com 
Tue Apr 14 00:56:28 CEST 2009

Hi expeRts,

I would like to calculate weighted mean by two factors.

My code is as follows:

R> tmp <- by(re$meta.sales.lkm[, c("pc", "sales")],
                      re$meta.sales.lkm[, c("size", "yr")], function(x)
                      weighted.mean(x[,1], x[,2]))

The result is as follows:
R> tmp
size: micro
yr: 1994
[1] 1.090
------------------------------------------------------------
size: small
yr: 1994
[1] 1.135
------------------------------------------------------------
size: medium
yr: 1994
[1] 1.113
------------------------------------------------------------
size: large
yr: 1994
[1] 1.105
------------------------------------------------------------
size: micro
yr: 1995
[1] 1.167
------------------------------------------------------------
size: small
yr: 1995
[1] 1.096
------------------------------------------------------------
size: medium
yr: 1995
[1] 1.056
....
....

But the form I want to get is as follows:
            1994       1995         1996      .....
micro    1.090      1.167         .............
small     1.135      1.096
medium 1.113      1.056        .... ........
large      1.105      ....... ...........

That is, the result should be tabularized.
How can I get the above form directly? (I don't want to modify tmp with
as.vector() and matrix() to get the result)

Thank  you in advance.

------------------------------------------------------------------------
--
Donghyun Oh
CESIS, KTH
------------------------------------------------------------------------
--

	[[alternative HTML version deleted]]