Hello,
I am working on fitting a logistic regression model to my dataset. I removed
the squared term in the second version of the model, but my model output is
exactly the same.
Model version 1:
GRP_GLM<-glm(HB_NHB~elev+costdis1^2,data=glm_1,family=binomial(link=logit))
summary(GRP_GLM)
Model version 2:
QM_1<-glm(HB_NHB~elev+costdis1,data=glm_2,family=binomial(link=logit))
summary(QM_1)
The call in version 2 has changed:
Call:
glm(formula = HB_NHB ~ elev + costdis1, family = binomial(link = logit),
data = glm_2)
But I’m getting the exact same results as I did in the model where costdis1 is
squared.
Any ideas what I might do to correct this? Thank you.
Sally
[[alternative HTML version deleted]]
On Nov 7, 2011, at 10:58 AM, Sally Ann Sims wrote:> Hello, > > I am working on fitting a logistic regression model to my dataset. > I removed the squared term in the second version of the model, but > my model output is exactly the same. > > Model version 1: GRP_GLM<-glm(HB_NHB~elev > +costdis1^2,data=glm_1,family=binomial(link=logit)) > summary(GRP_GLM) > > > Model version 2: QM_1<-glm(HB_NHB~elev > +costdis1,data=glm_2,family=binomial(link=logit)) > summary(QM_1) > > > The call in version 2 has changed: > Call: > glm(formula = HB_NHB ~ elev + costdis1, family = binomial(link = > logit), > data = glm_2) > But I???m getting the exact same results as I did in the model where > costdis1 is squared.Are you sure that you got output that correctly modeled the costdis1^2? I would ahve guessed that you would have needed to use : GRP_GLM<-glm(HB_NHB~elev+I(costdis1^2), data=glm_1, family=binomial(link=logit)) ?I The "^" in model formulas is for composing interactions. ?formula> > Any ideas what I might do to correct this? Thank you. > > Sally > [[alternative HTML version deleted]]And please post in plain text. -- David Winsemius, MD West Hartford, CT
Since you didn't provide a reproducible example, here are a couple of
possibilities to check, but I have utterly no idea if they're
applicable to your problem or not:
* does costdis1 consist of 0's and 1's?
* is costdis1 a factor?
In the first model, you treat costdis1 as a pure quadratic and in the
second model, it is a linear term. The two models are not nested.
Modeling a term as a pure quadratic is a very strong assumption - the
more usual practice is to fit both a linear and quadratic term in
costdis1 to allow more flexibility in the fitted surface, but that
would require costdis1 to be numeric.
HTH,
Dennis
On Mon, Nov 7, 2011 at 7:58 AM, Sally Ann Sims <sallysims at
earthlink.net> wrote:> Hello,
>
> I am working on fitting a logistic regression model to my dataset. ?I
removed the squared term in the second version of the model, but my model output
is exactly the same.
>
> Model version 1:
?GRP_GLM<-glm(HB_NHB~elev+costdis1^2,data=glm_1,family=binomial(link=logit))
> summary(GRP_GLM)
>
>
> Model version 2:
?QM_1<-glm(HB_NHB~elev+costdis1,data=glm_2,family=binomial(link=logit))
> summary(QM_1)
>
>
> The call in version 2 has changed:
> Call:
> glm(formula = HB_NHB ~ elev + costdis1, family = binomial(link = logit),
> ? ?data = glm_2)
> But I?m getting the exact same results as I did in the model where costdis1
is squared.
>
> Any ideas what I might do to correct this? ?Thank you.
>
> Sally
> ? ? ? ?[[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.
>
>