Re: [R] question about glm vs. loglin()

2011-09-17 Thread Ted Harding
On 17-Sep-11 01:20:53, Rolf Turner wrote: > On 17/09/11 01:19, Michael Friendly wrote: > > >> Trying to interpret associations in complex loglinear models >> from tables of parameter estimates is like trying to extract >> sunlight from a cucumber. You have to squeeze very hard, and >> then are

Re: [R] question about glm vs. loglin()

2011-09-16 Thread Rolf Turner
On 17/09/11 01:19, Michael Friendly wrote: Trying to interpret associations in complex loglinear models from tables of parameter estimates is like trying to extract sunlight from a cucumber. You have to squeeze very hard, and then are usually unhappy with the quality of the sunlight. For

Re: [R] question about glm vs. loglin()

2011-09-16 Thread Michael Friendly
Hi Yana Trying to interpret associations in complex loglinear models from tables of parameter estimates is like trying to extract sunlight from a cucumber. You have to squeeze very hard, and then are usually unhappy with the quality of the sunlight. Instead, you can visualize the association

Re: [R] question about glm vs. loglin()

2011-09-15 Thread David Winsemius
On Sep 15, 2011, at 4:33 PM, Yana Kane-Esrig wrote: Dear R gurus, I am looking for a way to fit a predictive model for a contingency table which has counts. I found that glm( family=poisson) is very good for figuring out which of several alternative models I should select. But once I sel

[R] question about glm vs. loglin()

2011-09-15 Thread Yana Kane-Esrig
Dear R gurus, I am looking for a way to fit a predictive model for a contingency table which has counts. I found that glm( family=poisson) is very good for figuring out which of several alternative models I should select. But once I select a model it is hard to present and interpret it, especia