ML Estimation Differences with R and SAS
On 11/03/2008, at 6:09 AM, Patrick Richardson wrote:
List, I'm working on fitting a logistic model for a well known dataset (which is given below in case anyone wants to try to reproduce). I used both R and SAS to fit the model and have some differences in the parameter estimates. I'm wondering if R calculates the ML estimates differently. I'm making NO accusations as to which program is "right or wrong". That is not the focus of this posting. As a "newer" R user I'm trying to understand the algorithm that R might use to calculate ML estimation. The largest difference seems to with the race factors. R gives a p-value of 0.46995 for race=black and SAS gives a p-value of 0.0753 for race=black. Clearly one is borderline significant and the other is not. Many thanks to all who might be able to offer any insight on this. Both R and SAS code and output are included in this message (along with the dataset).
Try setting
options(contrasts=c("contr.SAS","contr.poly"))
before you run your analysis in R.
cheers,
Rolf Turner
######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:9}}