r - Deviance of subset of data -

hi fit non linear curve , check deviance different segments of data, because curve behaves strangly fit data in different segments minimizes deviance @ cost of other segments.

is there way use deviance function subset. have 100 observations , want @ deviance @ bottom 50 , top 100

deviance(reg.model)[1:50], deviance(reg.model)[50:100] 

or can use deviance other data? otherwise have write own function sum of squared residuals.

i realized need use data.frame right variable name second argument predict. not sure if works deviance though.

predict(reg.model, data.frame(x.var=seq(1,10))  

would give out of sample predictions 1 10 of reg model