Goodmorning everybody, i'm an italian statistician and i'm using R for research. Could someone tell me some indices to see the goodness of fit in multilevel modelling? I'm using the lmer function, and I want to know if my model fit well my data. I actually want to justify the use of multilevel model instead the classical one. Hope someone can help me. Thank you. Greetings Chiara
As you have not provided a clue of what your models are, one can only guess. But if you mean using lots of fixed effects vs a random effect, the answer is that there is no such animal. They are two different non-nested models, and should be chosen based on subject matter considerations. Standard practice is to compare AIC, BIC and other "information criteria" but there is no clear standard to determine how large a difference is meaningful. You may wish to post this on R-SIG-MIXED-MODELS for other inputs. Bert Sent from my iPhone -- please excuse typos. On Apr 30, 2012, at 3:21 AM, "klair87 at libero.it" <klair87 at libero.it> wrote:> Goodmorning everybody, > i'm an italian statistician and i'm using R for research. > > Could someone tell me some indices to see the goodness of fit in multilevel > modelling? > I'm using the lmer function, and I want to know if my model fit well my > data. > I actually want to justify the use of multilevel model instead the classical > one. > > Hope someone can help me. > Thank you. > > Greetings > Chiara > > ______________________________________________ > R-help at r-project.org 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.
Hi Chiara, If you just want to compare model fit, you could use a LRT between models where you do and do not estimate the variance/covariance matrix of random effects. R^2 in mixed models do not have the same nice properties they do in fixed effects models. Cheers, Josh On Mon, Apr 30, 2012 at 3:21 AM, klair87 at libero.it <klair87 at libero.it> wrote:> Goodmorning everybody, > i'm an italian statistician and i'm using R for research. > > Could someone tell me some indices to see the goodness of fit in multilevel > modelling? > I'm using the lmer function, and I want to know if my model fit well my > data. > I actually want to justify the use of multilevel model instead the classical > one. > > Hope someone can help me. > Thank you. > > Greetings > Chiara > > ______________________________________________ > R-help at r-project.org 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.-- Joshua Wiley Ph.D. Student, Health Psychology Programmer Analyst II, Statistical Consulting Group University of California, Los Angeles https://joshuawiley.com/
If you want to know if your model fit will your data, then looking at residual-type plots is useful, as can be plotting the model-predictions and observed data together. You might also find interesting "R2 statistics for mixed models" by Matthew Kramer. Beware--as others have indicated, there is no widely-accepted measure of R2 for mixed models. Kevin On Mon, Apr 30, 2012 at 5:21 AM, klair87@libero.it <klair87@libero.it>wrote:> Goodmorning everybody, > i'm an italian statistician and i'm using R for research. > > Could someone tell me some indices to see the goodness of fit in multilevel > modelling? > I'm using the lmer function, and I want to know if my model fit well my > data. > I actually want to justify the use of multilevel model instead the > classical > one. > > Hope someone can help me. > Thank you. > > Greetings > Chiara > > ______________________________________________ > R-help@r-project.org 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. >-- Kevin Wright [[alternative HTML version deleted]]