[R-sig-ME] BLUPs in relation to fixed effect interaction
Luca Borger
lborger at cebc.cnrs.fr
Wed Jun 13 12:33:04 CEST 2012
Hello,
>We are interested in how much the villages deviate from the estimate
of the interaction Question*treat
not sure I got your design and question right, but why not including a
random slope (with the constraint that the covariate cannot be constant
within each level
of the grouping factor of the random effect)?
More generally, I wonder if psychometric methods and item response
theory might be of interest to you (apologies if I'm misunderstanding):
http://www.jstatsoft.org/v20/a02/paper
Cheers,
Luca
# Forthcoming book chapter
# Dispersal Ecology and Evolution (ch. 17)
# http://ukcatalogue.oup.com/product/9780199608904.do
---------------------------------------------------------------------
Luca Borger
Postdoctoral Research Fellow
Centre d'Etudes Biologiques de Chizé
CNRS (UPR1934); INRA (USC1339)
79360 Villiers-en-Bois, France
Tel: +33 (0)549 09 96 13
Fax: +33 (0)549 09 65 26
email: lborger at cebc.cnrs.fr
Web: http://cnrs.academia.edu/LucaBorger
Researcher ID: http://www.researcherid.com/rid/C-6003-2008
Google Scholar: http://scholar.google.com/citations?user=D5CTvNUAAAAJ
---------------------------------------------------------------------
# Newly published! Animal Migration: A synthesis (ch. 8):
# http://ukcatalogue.oup.com/product/9780199568994.do
Le 13/06/2012 11:33, Bunnefeld, Nils a écrit :
> Dear List,
>
>
> We are using an unmatched count technique to understand the prevalence of illegal hunting. This method is used to make sure people are anonymous and thus hopefully give honest answers about their illegal activity. People are allocated to two treatments and get either a card with four activities (teaching, farming, livestock herding and trading,) or with five activities (teaching, farming, livestock herding, trading and hunting of which one is the sensitive question). The respondents then indicate how many activities they are involved in on their card. We have also asked each person four times the same question but wanted to know whether each of the activities relate to the dry or wet season and whether it was for cash or non-cash. So we have Counts as a dependent variable as the number of activities each person does, Question with four levels (cash dry, cash wet, NonCash dry, NonCash wet), treat with two levels (card with 4 activities, card with 5 activities including the se!
> nsitive question, Treatment). We have run the model below with Village and id as random effects.
>
> We are now interested why villages are different from each other and extracted the random effects BLUPs using ranef() to be able to use village level explanatory variables in a new model (e.g. distance to protected area). This will give us the estimate how much each village deviates from the overall number of activities, the overall intercept; not really what we are interested in. We are interested in how much the villages deviate from the estimate of the interaction Question*treat because this gives us estimates about the prevalence of illegal hunting in the different seasons rather than the number of activities people are involved in. Any comments or ideas how to implement this in R would be greatly appreciated. An output from our model is below.
>
> Many thanks,
> Nils
>
>
>> library(lme4)
>
>> m1 <- lmer(Counts~Question*treat+(1|Village/id),data=data2,REML=F)
>
>
>
>> summary(m1)
>
> Linear mixed model fit by maximum likelihood
>
> Formula: Counts ~ Question * treat + (1 | Village/id)
>
> Data: data2
>
> AIC BIC logLik deviance REMLdev
>
> 7806 7876 -3892 7784 7825
>
> Random effects:
>
> Groups Name Variance Std.Dev.
>
> id:Village (Intercept) 0.21712 0.46596
>
> Village (Intercept) 0.13535 0.36791
>
> Residual 0.23434 0.48409
>
> Number of obs: 4356, groups: id:Village, 1092; Village, 15
>
>
>
> Fixed effects:
>
> Estimate Std. Error t value
>
> (Intercept) 1.488733 0.099243 15.001
>
> QuestionCash wet 0.095400 0.029285 3.258
>
> QuestionNonCash Dry 0.285618 0.029226 9.773
>
> QuestionNonCash Wet 0.514709 0.029226 17.611
>
> treatTreatment 0.184647 0.040999 4.504
>
> QuestionCash wet:treatTreatment -0.065976 0.041561 -1.587
>
> QuestionNonCash Dry:treatTreatment0.005799 0.041520 0.140
>
> QuestionNonCash Wet:treatTreatment-0.030882 0.041532 -0.744
>
>
>
>
>
>> ranef(m1)
>
> $Village
>
> (Intercept)
>
> Guta 0.0822752928
>
> Hunyari -0.0614857190
>
> Ketembere 0.6976740857
>
> Kitunguruma -0.3494539970
>
> Koreri -0.1217766001
>
> Kunzugu -0.4999783998
>
> Ligamba -0.4522420174
>
> Makundusi 0.8675942165
>
> Manyamanyama -0.1824419221
>
> Merenga -0.0006875998
>
> Migungani -0.1540653530
>
> Morotonga 0.1473949913
>
> Nyamburu -0.1751358164
>
> Nyamoko 0.0567038683
>
> Robanda 0.1456249700
>
>
>
>
>
> ------------------------------------------------------------------------
> Dr Nils Bunnefeld
> Imperial College London
> Silwood Park
> SL5 7PY, Ascot, UK
> http://www.iccs.org.uk/nils-bunnefeld
> http://fp7hunt.net/
> Tel: +44 20 7594 9086
>
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>
More information about the R-sig-mixed-models
mailing list