An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20090219/28d1cea2/attachment-0001.pl>
type III effect from glm()
5 messages · (Ted Harding), Simon Pickett, Mark Difford
On 19-Feb-09 10:38:50, Simon Pickett wrote:
Hi all,
This could be naivety/stupidity on my part rather than a problem
with model output, but here goes....
I have fitted a fairly simple model
m1<-glm(count~siteall+yrs+yrs:district,family=quasipoisson,
weights=weight,data=m[x[[i]],])
I want to know if yrs (a continuous variable) has a significant
unique effect in the model, so I fit a simplified model with the
main effect ommitted...
m2<-glm(count~siteall+yrs:district,family=quasipoisson,
weights=weight,data=m[x[[i]],])
So, above, you have fitted two models: m1, m2
then compare models using anova() anova(m1,m1b,test="F")
And here you are comparing two models: m1, m1b Could this be the reason for your result?
Analysis of Deviance Table Model 1: count ~ siteall + yrs + yrs:district Model 2: count ~ siteall + yrs:district Resid. Df Resid. Dev Df Deviance F Pr(>F) 1 1936 75913 2 1936 75913 0 0 The d.f.'s are exactly the same, is this right? Can I only test the significance of a main effect when it is not in an interaction? Thanks in advance, Simon.
-------------------------------------------------------------------- E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk> Fax-to-email: +44 (0)870 094 0861 Date: 19-Feb-09 Time: 10:56:12 ------------------------------ XFMail ------------------------------
Sorry, that was a typo in the email, not the model. So I still have the problem..... Cheers, Simon. ----- Original Message ----- From: "Ted Harding" <Ted.Harding at manchester.ac.uk> To: "Simon Pickett" <simon.pickett at bto.org>; <r-help at r-project.org> Sent: Thursday, February 19, 2009 10:56 AM Subject: RE: [R] type III effect from glm()
On 19-Feb-09 10:38:50, Simon Pickett wrote:
Hi all,
This could be naivety/stupidity on my part rather than a problem
with model output, but here goes....
I have fitted a fairly simple model
m1<-glm(count~siteall+yrs+yrs:district,family=quasipoisson,
weights=weight,data=m[x[[i]],])
I want to know if yrs (a continuous variable) has a significant
unique effect in the model, so I fit a simplified model with the
main effect ommitted...
m2<-glm(count~siteall+yrs:district,family=quasipoisson,
weights=weight,data=m[x[[i]],])
So, above, you have fitted two models: m1, m2
then compare models using anova() anova(m1,m2,test="F")
And here you are comparing two models: m1, m1b Could this be the reason for your result?
Analysis of Deviance Table Model 1: count ~ siteall + yrs + yrs:district Model 2: count ~ siteall + yrs:district Resid. Df Resid. Dev Df Deviance F Pr(>F) 1 1936 75913 2 1936 75913 0 0 The d.f.'s are exactly the same, is this right? Can I only test the significance of a main effect when it is not in an interaction? Thanks in advance, Simon.
-------------------------------------------------------------------- E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk> Fax-to-email: +44 (0)870 094 0861 Date: 19-Feb-09 Time: 10:56:12 ------------------------------ XFMail ------------------------------
Hi Simon,
I want to know if yrs (a continuous variable) has a significant unique effect in the model, so I fit a simplified model with the main effect ommitted...
[A different approach...] This is not really a sensible question until you have established that there is no significant interaction between "yrs" and "district." If this interaction is significant then, ipso facto, the effect of "yrs" is not unique but depends on "district." So establish that first. There is a good section on marginality in MASS (Venables & Ripley) and, as Mark has mentioned, in Prof Fox's texts. From what I can remember, some of these tests are reparametrized behind the scenes to enforce the marginality constraint. Regards, Mark.
Simon Pickett-4 wrote:
Hi all, This could be naivety/stupidity on my part rather than a problem with model output, but here goes.... I have fitted a fairly simple model m1<-glm(count~siteall+yrs+yrs:district,family=quasipoisson,weights=weight,data=m[x[[i]],]) I want to know if yrs (a continuous variable) has a significant unique effect in the model, so I fit a simplified model with the main effect ommitted... m2<-glm(count~siteall+yrs:district,family=quasipoisson,weights=weight,data=m[x[[i]],]) then compare models using anova() anova(m1,m1b,test="F") Analysis of Deviance Table Model 1: count ~ siteall + yrs + yrs:district Model 2: count ~ siteall + yrs:district Resid. Df Resid. Dev Df Deviance F Pr(>F) 1 1936 75913 2 1936 75913 0 0
The d.f.'s are exactly the same, is this right? Can I only test the significance of a main effect when it is not in an interaction? Thanks in advance, Simon. Dr. Simon Pickett Research Ecologist Land Use Department Terrestrial Unit British Trust for Ornithology The Nunnery Thetford Norfolk IP242PU 01842750050 [[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
View this message in context: http://www.nabble.com/type-III-effect-from-glm%28%29-tp22097773p22099331.html Sent from the R help mailing list archive at Nabble.com.
Hi Simon,
[On my response] ...not really a sensible question until...
On reading through this...what I mean is that yours seems not to be a "sensible approach," the question itself may be reasonable. What you want to be doing is testing whether the interaction term (yrs:district) gets dropped. Do it by comparing nested models (basically as you have done), or use dropterm() or stepAIC() [both are in MASS]. Regards, Mark.
Mark Difford wrote:
Hi Simon,
I want to know if yrs (a continuous variable) has a significant unique effect in the model, so I fit a simplified model with the main effect ommitted...
[A different approach...] This is not really a sensible question until you have established that there is no significant interaction between "yrs" and "district." If this interaction is significant then, ipso facto, the effect of "yrs" is not unique but depends on "district." So establish that first. There is a good section on marginality in MASS (Venables & Ripley) and, as Mark has mentioned, in Prof Fox's texts. From what I can remember, some of these tests are reparametrized behind the scenes to enforce the marginality constraint. Regards, Mark. Simon Pickett-4 wrote:
Hi all, This could be naivety/stupidity on my part rather than a problem with model output, but here goes.... I have fitted a fairly simple model m1<-glm(count~siteall+yrs+yrs:district,family=quasipoisson,weights=weight,data=m[x[[i]],]) I want to know if yrs (a continuous variable) has a significant unique effect in the model, so I fit a simplified model with the main effect ommitted... m2<-glm(count~siteall+yrs:district,family=quasipoisson,weights=weight,data=m[x[[i]],]) then compare models using anova() anova(m1,m1b,test="F") Analysis of Deviance Table Model 1: count ~ siteall + yrs + yrs:district Model 2: count ~ siteall + yrs:district Resid. Df Resid. Dev Df Deviance F Pr(>F) 1 1936 75913 2 1936 75913 0 0
The d.f.'s are exactly the same, is this right? Can I only test the significance of a main effect when it is not in an interaction? Thanks in advance, Simon. Dr. Simon Pickett Research Ecologist Land Use Department Terrestrial Unit British Trust for Ornithology The Nunnery Thetford Norfolk IP242PU 01842750050 [[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
View this message in context: http://www.nabble.com/type-III-effect-from-glm%28%29-tp22097773p22099812.html Sent from the R help mailing list archive at Nabble.com.