Re: [R] R^2 from lme function
On 5/14/07, Martin Henry H. Stevens <[EMAIL PROTECTED]> wrote: > Hi Cleber, > By "full" I simply meant "not REML." the function assumes that the > fixed effects were estimated using REML criteria, and using update() > simply changes that to ML. If the model was fit originally with ML, > it shouldn't make any difference. > I am reasonably sure that it should not matter whether there is an > intercept. ML estimates are invariant to fixed effects structure, > whereas REML depends upon it. I think the issue with presence or absence of an intercept is in how it affects the choice of the null model. Your function always uses y ~ 1 as the null model and that may not be appropriate if there is no intercept term in the original model. > On May 14, 2007, at 11:55 AM, Cleber Borges wrote: > > > Hi Martin, > > > > many thanks for your tip! > > > > but,{ :-( } > > what it 'full MLE' ? how to calculate? it is a saturated model??? > > > > and > > > > it is valid for 'no-intercept model? > > > > > > Many thanks again... > > > > Cleber > > > > > >> Hi Cleber, > >> I have been using this function I wrote for lmer output. It should be > >> easy to convert to lme. As with everything, buyer beware. Note > >> that it > >> requires (full) maximum likelihood estimates. > >> > >> > >> Rsq <- function(reml.mod) { > >> ## Based on > >> ## N. J. D. Nagelkerke. A note on a general definition > >> ## of the coefficient of determination. Biometrika, 78:691–692, > >> 1991. > >> ml.mod <- update(reml.mod, method="ML") > >> l.B <- logLik(ml.mod) > >> l.0 <- logLik( lm([EMAIL PROTECTED] ~ 1) ) > >> Rsq <- 1 - exp( - ( 2/length([EMAIL PROTECTED]) ) * (l.B - l.0) ) > >> Rsq[1] > >> } > >> > >> Hank > >> > >> > >> > >> > >>> Hello allR > >>> How to access R^2 from lme object? > >>> or how to calculate it? > >>> ( one detail: my model do not have a intercept ) > >>> thanks in advanced > >>> Cleber > > > > > > ___ > > > > Experimente já e veja as novidades. > > > > __ > > R-help@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. > > > > Dr. Hank Stevens, Assistant Professor > 338 Pearson Hall > Botany Department > Miami University > Oxford, OH 45056 > > Office: (513) 529-4206 > Lab: (513) 529-4262 > FAX: (513) 529-4243 > http://www.cas.muohio.edu/~stevenmh/ > http://www.muohio.edu/ecology/ > http://www.muohio.edu/botany/ > > "E Pluribus Unum" > > __ > R-help@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. > __ R-help@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.
Re: [R] R^2 from lme function
Hi Cleber, By "full" I simply meant "not REML." the function assumes that the fixed effects were estimated using REML criteria, and using update() simply changes that to ML. If the model was fit originally with ML, it shouldn't make any difference. I am reasonably sure that it should not matter whether there is an intercept. ML estimates are invariant to fixed effects structure, whereas REML depends upon it. Cheers, Hank On May 14, 2007, at 11:55 AM, Cleber Borges wrote: > Hi Martin, > > many thanks for your tip! > > but,{ :-( } > what it 'full MLE' ? how to calculate? it is a saturated model??? > > and > > it is valid for 'no-intercept model? > > > Many thanks again... > > Cleber > > >> Hi Cleber, >> I have been using this function I wrote for lmer output. It should be >> easy to convert to lme. As with everything, buyer beware. Note >> that it >> requires (full) maximum likelihood estimates. >> >> >> Rsq <- function(reml.mod) { >> ## Based on >> ## N. J. D. Nagelkerke. A note on a general definition >> ## of the coefficient of determination. Biometrika, 78:691–692, >> 1991. >> ml.mod <- update(reml.mod, method="ML") >> l.B <- logLik(ml.mod) >> l.0 <- logLik( lm([EMAIL PROTECTED] ~ 1) ) >> Rsq <- 1 - exp( - ( 2/length([EMAIL PROTECTED]) ) * (l.B - l.0) ) >> Rsq[1] >> } >> >> Hank >> >> >> >> >>> Hello allR >>> How to access R^2 from lme object? >>> or how to calculate it? >>> ( one detail: my model do not have a intercept ) >>> thanks in advanced >>> Cleber > > > ___ > > Experimente já e veja as novidades. > > __ > R-help@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. Dr. Hank Stevens, Assistant Professor 338 Pearson Hall Botany Department Miami University Oxford, OH 45056 Office: (513) 529-4206 Lab: (513) 529-4262 FAX: (513) 529-4243 http://www.cas.muohio.edu/~stevenmh/ http://www.muohio.edu/ecology/ http://www.muohio.edu/botany/ "E Pluribus Unum" __ R-help@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.
Re: [R] R^2 from lme function
Hi Martin, many thanks for your tip! but,{ :-( } what it 'full MLE' ? how to calculate? it is a saturated model??? and it is valid for 'no-intercept model? Many thanks again... Cleber > Hi Cleber, > I have been using this function I wrote for lmer output. It should be > easy to convert to lme. As with everything, buyer beware. Note that it > requires (full) maximum likelihood estimates. > > > Rsq <- function(reml.mod) { > ## Based on > ## N. J. D. Nagelkerke. A note on a general definition > ## of the coefficient of determination. Biometrika, 78:691–692, 1991. > ml.mod <- update(reml.mod, method="ML") > l.B <- logLik(ml.mod) > l.0 <- logLik( lm([EMAIL PROTECTED] ~ 1) ) > Rsq <- 1 - exp( - ( 2/length([EMAIL PROTECTED]) ) * (l.B - l.0) ) > Rsq[1] > } > > Hank > > > > >> Hello allR >> How to access R^2 from lme object? >> or how to calculate it? >> ( one detail: my model do not have a intercept ) >> thanks in advanced >> Cleber ___ Experimente já e veja as novidades. __ R-help@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.
Re: [R] R^2 from lme function
Hi Cleber, I have been using this function I wrote for lmer output. It should be easy to convert to lme. As with everything, buyer beware. Note that it requires (full) maximum likelihood estimates. Rsq <- function(reml.mod) { ## Based on ## N. J. D. Nagelkerke. A note on a general definition ## of the coefficient of determination. Biometrika, 78:691–692, 1991. ml.mod <- update(reml.mod, method="ML") l.B <- logLik(ml.mod) l.0 <- logLik( lm([EMAIL PROTECTED] ~ 1) ) Rsq <- 1 - exp( - ( 2/length([EMAIL PROTECTED]) ) * (l.B - l.0) ) Rsq[1] } Hank On May 14, 2007, at 10:50 AM, Cleber Borges wrote: > Hello allR > > > How to access R^2 from lme object? > or how to calculate it? > > ( one detail: my model do not have a intercept ) > > > thanks in advanced > > Cleber > > > > > > ___ > > Experimente já e veja as novidades. > > __ > R-help@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. Dr. Hank Stevens, Assistant Professor 338 Pearson Hall Botany Department Miami University Oxford, OH 45056 Office: (513) 529-4206 Lab: (513) 529-4262 FAX: (513) 529-4243 http://www.cas.muohio.edu/~stevenmh/ http://www.muohio.edu/ecology/ http://www.muohio.edu/botany/ "E Pluribus Unum" __ R-help@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.