[R] misclassification rate

[Patrick Breheny]

On 09/02/2011 05:29 PM, Doussa wrote:
> I have a matrix in wich there are two colmns( yp, yt)
> Yp: predicted values from my model.
> yt: true values ( my dependante variable y is a categorical;3 modalities
> (0,1,2)
> I don't know how to procede to calculate the misclassification rate and the
> error Types.

Suppose your data looks like this:

 > yp <- sample(0:2,50,replace=TRUE)
 > yt <- sample(0:2,50,replace=TRUE)

You can create a cross-classification table with:

 > tab <- table(yp,yt)
 > tab
    yt
yp   0  1  2
   0  5  8  5
   1  2 11  9
   2  1  2  7

And the misclassification rate is

 > 1-sum(diag(tab))/sum(tab)
[1] 0.54

-- 
Patrick Breheny
Assistant Professor
Department of Biostatistics
Department of Statistics
University of Kentucky