![]() ![]() Italian Institute of Technology email: gabriel.baud-bovy using iit.itĬentral office: via Morego 30, 16163 Genova tel.: (+39) 010 8172 202Įrzelli office: via Melen 73, 16152 Genova mobile: (+39) 3 > anova(fitRespLat11,fitRespLat12,fitRespLat13,fitRespLat14)įitRespLat14: resp.lat ~ stake.i.c + dir.c + + (1 | su)įitRespLat13: resp.lat ~ stake.i.c + dir.c + + (stake.i.c || su)įitRespLat12: resp.lat ~ stake.i.c + dir.c + + (stake.i.c + dir.c || su)įitRespLat11: resp.lat ~ stake.i.c + dir.c + + (stake.i.c + dir.c + || su)ĭf AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq)įitRespLat13 7 58129 58173 -29057 58115 76.699 1 ![]() Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) įormula: resp.lat ~ stake.i.c + dir.c + + (stake.i.c + dir.c + || su)Ĭontrol: glmerControl(optimizer = "bobyqa") If anybody is interested, I might share the With identity link and welcome other suggestion but I would also like to understand what As I have mentioned in a previous question to the list, my general goal is to be able to fit RT I also don't understand the value of residual SD, which seems to be on a different scale Would be more trustworthy with gamma noise than with gaussian noise. I would expect therefore that the result of the model It is a better model than Gaussian noise. Even if gamma distribution is not a perfect model of RT variability (because of underdispersion), Intuitively, I would expect thatĮstimates to be not statistically significant if there is a lot of unexplained variability. I don't understand why the results are statistically significant. Note that the model fits without warning. The randomĮffects also appear to be needed (statistically significant LRTs) and supported by the data (rePCA)Īlthough I double that it is the case. For example, all fixed effect are statistically significantĭespite the fact that these factors explain little to nothing when looking at the plots. The results of the model don't make sense. #Gamma control ic how toThe DHARMa residual plot suggests underdispersion and I don't know how to deal with that. (otherwise, I'll cross post the answers). I have put some info at the end of this email it is probably best to answer me on stakeexchange, where it is also possible to see the plots I mention this question in the mailing list because there are still things that I don't understand. I have updated a one-month old post on stakexchange about a GLMM model that I used to fit RT with gamma Next message (by thread): R.square in Mixed Models.Previous message (by thread): Fixing singularity in a generalized linear mixed effect model.Fitting RT: underdispersion with gamma and identity link Baud-Bovy Gabriel |rom Mar 21 14:06: Fitting RT: underdispersion with gamma and identity link ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |