Leave one out cross validation

#1
Hi everybody,
I've done a linear regression analysis with forward selection method. my number of observations are 32 and the final model selected 2 independents out of 27 variables entered to the regression analysis. now, I am asked to do a leave one out cross validation for checking the predictivity of the model. but I know a few about statistics and LOOCV. I wonder if it is possible to do LOOCV in SPSS and if it is possible, how should I do that.
Regards
 
Last edited:

rogojel

TS Contributor
#2
hi,
I dont lnow SPSS but I suspect that it is already in the output somewhere. E.g. in Minitab it is called PRESS , predictive sum of squares. Maybe you will find something similar.

regards
 
#3
Thanks a lot. I saw leave one out cross validation option in Partial Least Square regression and PRESS value in Minitab. but is it possible to do cross validation for multiple linear regression in Minitab? and how should I interprete value of PRESS as an index of model validity?
Regards
 

rogojel

TS Contributor
#4
Hi,
yes, you just need to check the option in Minitab, in the Outputs dialog box, if memory serves.

Press actually means predicred sum if squares so the smaller the better.

regards