I didn't know that! Now all makes sense. I thought the value corresponded to the actual estimate of the parameter. So, for the intercerpt it's the same?
El sáb., 12 de oct. de 2019 a la(s) 04:54, HADFIELD Jarrod ( j.hadfi...@ed.ac.uk) escribió: > Hi, > > The random slopes are deviations. You need to add the fixed effect > associated with x to get the actual slopes. Does this deal with your > reservations? > > Cheers, > > Jarrod > > On 12 Oct 2019, at 08:18, Oscar Inostroza <oskinostr...@gmail.com> wrote: > > Hi, > thank you for your attention. My data consists of 1820 observations > (species) grouped in 11 levels (I know that it is a lot!). I mean "crazy" > estimates because if I run the models separately for each level (i.e. > trimmed the database at each level), there is a very strong positive > relationship between x and y. However, when running the full model, the > slopes and intercepts are very different (e.g. negative or null slopes). > > El sáb., 12 de oct. de 2019 a la(s) 03:51, HADFIELD Jarrod ( > j.hadfi...@ed.ac.uk) escribió: > >> Hi Oscar, >> >> The syntax and prior look fine. How many levels of group are there and >> what do you mean by crazy? >> >> Cheers, >> >> Jarrod >> >> >> > On 12 Oct 2019, at 04:54, Oscar Inostroza <oskinostr...@gmail.com> >> wrote: >> > >> > I'm trying to run a mixed phylogenetic model using the function in the >> > MCMCglmm R package of the same name. In this model, I want to estimate >> > slopes and intercepts for each level of a given variable (modeled as a >> > random effect). I tried to run the model with the following code. >> However, >> > the estimates of slopes and intercepts for the random effect are crazy! >> > (i.e. when running separate models for each level, the results are >> > reasonable) There is something wrong with my code? (in the syntax of the >> > model, or likely in the prior specification) >> > >> > prior<- list(G=list(G1=list(V=1, nu=0.002), G2=list(V=diag(2), nu=2, >> > alpha.mu= rep(0,2), alpha.V=diag(1000,2,2))), R=list(V=1, nu=0.002)) >> > >> > model<-MCMCglmm(y~x, random=~phylo + >> > >> us(1+x):group,family="gaussian",ginverse=list(phylo=inv.phylo$Ainv),prior=prior,data=data,nitt=nitt,burnin=burn,thin=thin, >> > pr=TRUE) >> > >> > Cheers >> > >> > p.s the same model, but in the syntax used by brms has the form: >> > >> > model <- brm(y ~ x +(1|phylo)+(1+x|group), cov_ranef = list(phylo = >> A)... >> > >> > (But it's too slow for my amount of data) >> > >> > -- >> > Oscar >> > >> > [[alternative HTML version deleted]] >> > >> > _______________________________________________ >> > R-sig-phylo mailing list - R-sig-phylo@r-project.org >> > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >> > Searchable archive at >> http://www.mail-archive.com/r-sig-phylo@r-project.org/ >> >> The University of Edinburgh is a charitable body, registered in Scotland, >> with registration number SC005336. >> > > > -- > Oscar Inostroza Michael > Biólogo > Magíster Ciencias (m) Zoología > Programa Doctorado en Sistemática y Biodiversidad > Laboratorio de Ecología Evolutiva y Filoinformática > Facultad de Ciencias Naturales Y Oceanográficas > Universidad de Concepción > Barrio Universitario s/n, Casilla 160-C > Concepción, CHILE > > Fono Oficina: (56) 41 - 220 – 7341 > > e-mail: oskinostr...@gmail.com > oscarinostr...@udec.cl > > Pagina Web: http://phyloevolecol.cl/inostroza-michael/ > > > -- Oscar Inostroza Michael Biólogo Magíster Ciencias (m) Zoología Programa Doctorado en Sistemática y Biodiversidad Laboratorio de Ecología Evolutiva y Filoinformática Facultad de Ciencias Naturales Y Oceanográficas Universidad de Concepción Barrio Universitario s/n, Casilla 160-C Concepción, CHILE Fono Oficina: (56) 41 - 220 – 7341 e-mail: oskinostr...@gmail.com oscarinostr...@udec.cl Pagina Web: http://phyloevolecol.cl/inostroza-michael/ [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
