R help - Nov 2012 - add1() alternative

Hi,

I'm trying to build a hierarchical logistic regression model with lme4
package, but I have a problem on selecting the variables to include in this
model.

In a simple logistic regression, using Forward selection, i use a likelihood
ratio test to check which variables i should include in the model, using the
function add1().

The problem is that this function doesn't work with the hierarchical model
that i'm trying to achieve.

Example:

model<- glmer(y ~ (1 | group)+x+sex+age, family =
binomial("logit"), data db)
add1(model, db, test="Chisq")
Error: $ operator not defined for this S4 class

I know that the drop1() function works fine to do a backward selection, but
the problem is that i have 40 variables, and which time i use drop1() i have
to wait a loooooong time to get a result...

Is there any alternative to add1()? I mean, i don't want to use
anova(model1,model2) because i would have to do all the models by hand...


Thanks a lot guys,

Bruno



--
View this message in context:
http://r.789695.n4.nabble.com/add1-alternative-tp4648215.html
Sent from the R help mailing list archive at Nabble.com.

Hi Bruno,

probably not exactly what you are looking for, but maybe "all subset
regression" as in library "leaps" might be an alternative for
variable
selection? But I am definitely not sure if this is faster than drop1()
(calculates more models), nor have I ever tested it with a hierarchical
logistic regression model.

Another option might be stepAIC() in MASS library, which is capable of
forward, backward and stepwise model selection. So it is add() and
drop() in one function.

Hope this might help a little.
Greetings from Munich, Felix


Am 02.11.12 10:49, schrieb brunosm:
> Hi,
>
> I'm trying to build a hierarchical logistic regression model with lme4
> package, but I have a problem on selecting the variables to include in this
> model.
>
> In a simple logistic regression, using Forward selection, i use a
likelihood
> ratio test to check which variables i should include in the model, using
the
> function add1().
>
> The problem is that this function doesn't work with the hierarchical
model
> that i'm trying to achieve.
>
> Example:
>
> model<- glmer(y ~ (1 | group)+x+sex+age, family =
binomial("logit"), data > db)
> add1(model, db, test="Chisq")
> Error: $ operator not defined for this S4 class
>
> I know that the drop1() function works fine to do a backward selection, but
> the problem is that i have 40 variables, and which time i use drop1() i
have
> to wait a loooooong time to get a result...
>
> Is there any alternative to add1()? I mean, i don't want to use
> anova(model1,model2) because i would have to do all the models by hand...
>
>
> Thanks a lot guys,
>
> Bruno
>
>
>
> --
> View this message in context:
http://r.789695.n4.nabble.com/add1-alternative-tp4648215.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help@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.

	[[alternative HTML version deleted]]