jackknife

#1
HI all, can some body pls let me know how can i do a jackknife regression. Actually i have done stepwise regression analysis and now i have to apply a jackknife method inorder to see if my regression analysis was robust or not. Further is it possible to do in SPSS....THANKS
 

Dragan

Super Moderator
#2
HI all, can some body pls let me know how can i do a jackknife regression. Actually i have done stepwise regression analysis and now i have to apply a jackknife method inorder to see if my regression analysis was robust or not. Further is it possible to do in SPSS....THANKS
I would use S-PLUS.
 
#3
Thanks and if i get correctly in case of jackknife we do the same regression (forexample, stepwise in my case) but the difference is only that we do it with one variable less each time. and then compare the results....

am i right........Regards to all
 

TheEcologist

Global Moderator
#4
Thanks and if i get correctly in case of jackknife we do the same regression (forexample, stepwise in my case) but the difference is only that we do it with one variable less each time. and then compare the results....

am i right........Regards to all
That seems to be the jackknife in a nutshell yes.
Try using the free software R (www.r-project.org).
It should be rather elementry code wise
(here's an example code because I'm such a nice guy :p ):

resamples=10
dataset=data.frame(x=seq(1,100,1),y=rnorm(100,2,5)+seq(1,100,1))
n=dim(dataset)[1]
results=data.frame(a=rep(NA,resamples),
b=rep(NA,resamples))

for(i in 1:resamples)
{
jackknife=dataset[sample(1:n,n-1),]
results[i,]=lm(y~x,data=jackknife)$coef
}

#saved coef for all jackknife regressions
results


PDF:
http://jaqm.ro/issues/volume-2,issue-2/pdfs/sahinler_topuz.pdf

R-code that might help:
http://bm2.genes.nig.ac.jp/RGM2/R_current/library/pls/R/jackknife.R