[R] Create subsets of data

[David Winsemius]

On May 9, 2011, at 3:46 PM, David Winsemius wrote:

>
> On May 9, 2011, at 9:40 AM, Pavan G wrote:
>
>> Hello All,
>> Let's say I have points on a x-y plane. x ranges from 0-2 and y  
>> from 0-2.
>> There are points in quadrants x[0:1]---y[0:1] and in x[1:2]---- 
>> y[1:2]. I
>> would like to get the mean and std of the points in the x[0:1]---- 
>> y[0:1]
>> quadrant alone. Is there a straight forward way to do it?
>
> I am assuming that you have yet a third vector that has values at  
> each of those coordinates and it is that set of categorized values  
> for which you want the means and std deviations: I will exemplify  
> the use of tapply on cut()-igorized vectors and a z-value which is  
> the L1 distance from the origin.

(Actually, after checking help(dist)  I think the sum(x_i) is called  
the Canberra distance. )

> (And excuse the earlier feeble attempt at humor);
>
> > x <- runif(2000, 0,2)
> > y <- runif(2000, 0, 2)
> > xc <- cut(x, c(0,1,2))
> > yc <- cut(y, c(0,1,2))
> > z <- x+y
> > tapply(z, list(xc,yc), mean)
>         (0,1]    (1,2]
> (0,1] 1.013192 2.016095
> (1,2] 1.979930 2.996229  # seems to make sense
> > tapply(z, list(xc,yc), sd)
>          (0,1]     (1,2]
> (0,1] 0.4028310 0.4133113
> (1,2] 0.4239014 0.3984559 # also seems sensible
> -- 



David Winsemius, MD
West Hartford, CT