Message-ID: <40e66e0b0612190650h5431ddd3yf60dfb92e1fafdf6@mail.gmail.com>
Date: 2006-12-19T14:50:12Z
From: Douglas Bates
Subject: A question on lmer() function
In-Reply-To: <5A3E65F71DFAAE41AC2D20B61B177C5ED0F991@CDU-MAIL.cdu-staff.local>
On 12/17/06, Guojing Yang <Guojing.Yang at cdu.edu.au> wrote:
> Dear R users,
> We have encountered a slight problem when using the lmer()
> function:
>
> 1. Data description: 11 locations; Nt: monthly mosquito population
> density from 1994-2005 in each location.
> 2. Question: to examine the degree of spatial heterogeneity in the
> system by testing model support for single versus multiple intercepts
> and slopes for the location effect. We applied the lmer() function and
> introduced location as spatial effect. Four different models were
> generated as following:
> a. 1 intercept & 1 slope : lmer(r~log(N)+(1|location)), where r <-
> log(Nt/Nt-1) and location is a random effect.
> b. Multi-intercept & 1 slope: lmer(r~location+log(N)+(1|location))
> c. 1 slope & multi-intercept: ??????
> d. Multi-intercept & multi-slope:
> lmer(r~location*log(N)+(1|location))
I'm not exactly sure what you want to do here but you definitely
should not have location as a fixed-effects term and a random effects
term of the form (1|location). Such terms would be confounded.
I think the models you want to fit would all have an (implicit)
intercept and a log(N) term in the fixed effects. They would differ
in terms of whether you have a random effect for the intercept by
location and/or a random effect for the slope with respect to log(N)
by location. Four variants are
# random effects for intercept only
lmer(r ~ log(N) + (1|location))
# random effects for slope only
lmer(r ~ log(N) + (0+log(N)|location))
# possibly correlated random effects for intercept and slope
lmer(r ~ log(N) + (log(N)|location))
# independent random effects for intercept and slope
lmer(r ~ log(N) + (1|location) + (0+log(N)|location))
> As you can see, we have had trouble defining model c. Does anyone know
> how to code this using lmer? We really appreciate any response.
> Guojing
>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at stat.math.ethz.ch 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.
>