Displaying 1 result from an estimated 1 matches for "sums1".
Did you mean:
sums
2011 May 26
0
'constrained' negative.binomial model estimates
...rigin-0, data=migration, family=quasipoisson())
where origin is a factor of origins (1 - n). This has the effect of
'constraining' the predicted flows to reproduce the observed outgoing
flows from each origin:
> sums <- aggregate(migration$flows, by=list(migration$origin), sum)
> sums1 <- aggregate(predict(model1, type='response'), by=list(migration$origin), sum)
> sums2 <- aggregate(predict(model2, type='response'), by=list(migration$origin), sum)
which works fine for both poisson and quasipoisson models. However, I
would also like to fit a negative.bin...