Message-ID: <40e66e0b0809111215t63e59548yb0d772dd59b0dd41@mail.gmail.com>
Date: 2008-09-11T19:15:03Z
From: Douglas Bates
Subject: models with no fixed effects
In-Reply-To: <48C92526020000B600039CA4@ZGRW38.uwcm.ac.uk>
On Thu, Sep 11, 2008 at 8:03 AM, Daniel Farewell
<farewelld at cardiff.ac.uk> wrote:
> I'm running into an error when using lmer to fit models with no fixed effects terms.
>
> For example, generating some data with
>
> df$y <- with(df <- data.frame(i = gl(5, 5), b = rep(rnorm(5), each = 5)), b + rnorm(25))
>
> and fitting like this
>
> fit1 <- lmer(y ~ 1 + (1 | i), df)
>
> works fine. But fitting like this
>
> fit0 <- lmer(y ~ 0 + (1 | i), df)
>
> gives the following error:
>
> CHOLMOD error: Pl?
> Error in mer_finalize(ans) :
> Cholmod error `invalid xtype' at file:../Cholesky/cholmod_solve.c, line 971
Admittedly that is a rather obscure error message. It is related to
the fact, apparently not verified, that we should have p, the number
of fixed-effects, greater than zero.
I should definitely add a check on p to the validate method. (In some
ways I'm surprised that it got as far as mer_finalize before kicking
an error). I suppose that p = 0 could be allowed and I could add some
conditional code in the appropriate places but does it really make
sense to have p = 0? The random effects are defined to have mean
zero. If you have p = 0 that means that E[Y] = 0. I would have
difficulty imagining when I would want to make that restriction.
Let me make this offer - if someone could suggest circumstances in
which such a model would make sense, I will add the appropriate
conditional code to allow for p = 0. For the time being I will just
add a requirement of p > 0 to the validate method.