Message-ID: <500FBCC6.30301@xtra.co.nz>
Date: 2012-07-25T09:30:46Z
From: Rolf Turner
Subject: error in running glm() function
In-Reply-To: <1343201292178-4637727.post@n4.nabble.com>
Ummmm, doesn't the error message tend to indicate to you that
at least one of the factors in the model has only one level?
That would make that factor a constant, whence it could not possibly
contribute any predictive power, whence it should be eliminated from
the model.
Note that you *must* have meant "glm()" not "lm()" --- the "family"
argument makes no sense with lm().
It is quite possible that your model could/should be written
(more simply --- understatement of the millennium!!!) as:
ml.logit <- glm(default ~ ., family=binomial,data=dev)
(assuming the given list of variates exhausts the columns of "dev").
Note that you do not need to specify link="logit" (or even link=logit;
the quotes are unnecessary!) since logit is the default link for the
binomial family.
cheers,
Rolf Turner
On 25/07/12 19:28, shrutibansal wrote:
> m1.logit <-lm(default ~
> amt.fac +
> age.fac +
> duration +
> chk_acct +
> history +
> purpose +
> sav_acct +
> employment +
> install_rate +
> pstatus +
> other_debtor +
> time_resid +
> property +
> other_install +
> housing +
> other_credits +
> job +
> num_depend +
> telephone +
> foreign
> , family = binomial(link = "logit"), data
> = dev)
>
> on running this code I am getting the following error
>
> Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) :
> contrasts can be applied only to factors with 2 or more levels
> ..
> please help me solving this problem.