search for: f3p

Displaying 2 results from an estimated 2 matches for "f3p".

Did you mean: f32
2012 Sep 25
1
REML - quasipoisson
...s=TRUE)$val ldsq<-sum(log(eigSq[1:16])) F3q<-(ldhsq-ldsq)/2 #4th term Mp=3 F4q<-Mp/2*log(2*pi*phiq) F1q+F3q-F4q m3$gcv #quite different #but if phiq is replaced by the Pearson estimate of the scale P<-sum((y-fitted(m3))^2/fitted(m3)) phip<-P/(500-sum(m3$edf)) F1p<-Dpq/(2*phip) F3p<-F3q #third term independent of scale F4p<-Mp/2*log(2*pi*phip) F1p+F3p-F4p m3$gcv #closer but still wrong #is there a value of phi which makes this work? f<-function(t) (Dpq/(2*t) + F3q + Mp/2*log(2*pi*t) - m3$gcv)^2 optimise(f,interval=c(0.5,1.5))$min->phix Dpq/(2*phix) + F3q + Mp/2...
2012 Oct 01
0
[Fwd: REML - quasipoisson]
...> #4th term > Mp=3 > F4q<-Mp/2*log(2*pi*phiq) > > F1q+F3q-F4q > m3$gcv > > #quite different > > #but if phiq is replaced by the Pearson estimate of the scale > P<-sum((y-fitted(m3))^2/fitted(m3)) > phip<-P/(500-sum(m3$edf)) > F1p<-Dpq/(2*phip) > F3p<-F3q > #third term independent of scale > F4p<-Mp/2*log(2*pi*phip) > F1p+F3p-F4p > m3$gcv > > #closer but still wrong > > #is there a value of phi which makes this work? > f<-function(t) (Dpq/(2*t) + F3q + Mp/2*log(2*pi*t) - m3$gcv)^2 > optimise(f,interval=c(0...