Question

Aloha,

I am trying to cast() a dataset in which every unique combination of W-X-Y returns the max number of Z AND the associated week. For example:

W       X       Y       week    Z  
w1      x1      y1      1       0  
w1      x1      y1      2      0.1  
w1      x1      y1      3      0.2  
w2      x2      y1      1      0.5   
w2      x2      y1      2      0.7  
w2      x2      y1      3      0.3  
w3      x1      y1      1      0.1  
w3      x1      y1      2      0.2  
w3      x1      y1      3      0.5  
w4      x2      y2      1      0.7  
w4      x2      y2      2      0.3  
w4      x2      y2      3      0.1  
w5      x1      y2      1      0.3  
w5      x1      y2      2      0.1  
w5      x1      y2      3      0.2  

Can I do this w/cast()? I am able to return just the max number of Z per unique W-X-Y combination, but not the week with the following:

cast(foo, W + X + Y ~ ., max, value="Z")

For the above dataset, I would like the output to look as such:

W       X       Y       week    Z
w1      x1      y1      3      0.2
w2      x2      y1      2      0.7 
w3      x1      y1      3      0.5 
w4      x2      y2      1      0.7 
w5      x1      y2      1      0.3  

Mahalo for your suggestions!

Answer 1

cast is not the right tool for this. Consider instead the functions in the plyr package:

library("plyr")
ddply(foo, .(W, X, Y), summarise, week=week[which.max(Z)], Z=max(Z))