Bert: Not clear to me. Where mentioned are the functions similar to glm, if you please? Bruce ______________> On May 11, 2017, at 8:39 AM, Bert Gunter <bgunter.4567 at gmail.com> wrote: > > ?glmulti seems clear enough to me. If not, pls check the reference given therein. > > Bert > >> On May 11, 2017 5:22 AM, "BR_email" <br at dmstat1.com> wrote: >> Thanks, Bert. I would expect the list to include, at least lm. The reference states, "See Examples section." >> But, there is nothing in that section or elsewhere!! >> Bruce >> >> Bert Gunter wrote: >>> Probably? : >>> >>> All functions for which a link function of the response is modeled as a linear predictor of the covariates, but the response need not be in the exponential family? Such a list of course cannot be "listed". I would expect the package documentation, especially vignettes, explains this in any case. Pls check. >>> >>> Hopefully, you'll get a more authoritative response if I'm wrong. >>> >>> Bert >>> >>> >>> >>> >>> >>> >>> >>> On May 11, 2017 4:15 AM, "Bruce Ratner PhD" <br at dmstat1.com <mailto:br at dmstat1.com>> wrote: >>> >>> R-helpers: >>> In the "glmulti" package, it states parameter fitfunction assumes >>> functions similar to glm, but doesn't list them. >>> What are the functions similar to glm that can be used with glmulti? >>> Bruce >>> >>> ______________________________________________ >>> R-help at r-project.org <mailto:R-help at r-project.org> mailing list -- >>> To UNSUBSCRIBE and more, see >>> https://stat.ethz.ch/mailman/listinfo/r-help >>> <https://stat.ethz.ch/mailman/listinfo/r-help> >>> PLEASE do read the posting guide >>> http://www.R-project.org/posting-guide.html >>> <http://www.R-project.org/posting-guide.html> >>> and provide commented, minimal, self-contained, reproducible code. >>> >>> >>[[alternative HTML version deleted]]
> On May 11, 2017, at 6:31 AM, Bruce Ratner PhD <br at dmstat1.com> wrote: > > Bert: > Not clear to me. > Where mentioned are the functions similar to glm, if you please?The basis for the similarity was stated as having an available link function (and I suspected, an inverse as well.) I, for one, wouldn't have been surprised if `lm` were not in the list because `glm` with a 'gaussian' link would provide the same capabilities. I read the basis as stating that a family-object be available (an object with the features described on the `?family` page. If I were correct, then running `methods(family)` would provide a list of the family objects that are available for loaded packages:> methods("family")[1] family.glm* family.lm* family.negbin* see '?methods' for accessing help and source code So lm would qualify as well.> require(lme4)Loading required package: lme4 Loading required package: Matrix> methods("family")[1] family.glm* family.glmResp* family.lm* family.lmResp* family.merMod* [6] family.negbin* family.nlsResp* see '?methods' for accessing help and source code Following Bert's advice (and so reading the manual for you), I find that Vincent Calcagno has stated the criterion somewhat differently in "Using glmulti with any type of statistical model, with examples." #-------- glmulti works outof-the-box with several types of function (such as lm, glm or coxph), but it can in principle be used with any such function `myttingfunction`, as long as 1. The function receives a model specication in the form of a formula; 2. The function ts the model by maximum likelihood, which can be accessed through the standard `LogLik` function; Even when the two conditions above are veried, complications arise because, unfortunately, dierent tting functions have dierent conventions regarding how characteristics of the t should be accessed. Indeed, most of them come from dierent packages with dierent authors and there is no common standard so far. #-------- -- David> Bruce > > ______________ > > > > >> On May 11, 2017, at 8:39 AM, Bert Gunter <bgunter.4567 at gmail.com> wrote: >> >> ?glmulti seems clear enough to me. If not, pls check the reference given therein. >> >> Bert >> >>> On May 11, 2017 5:22 AM, "BR_email" <br at dmstat1.com> wrote: >>> Thanks, Bert. I would expect the list to include, at least lm. The reference states, "See Examples section." >>> But, there is nothing in that section or elsewhere!! >>> Bruce >>> >>> Bert Gunter wrote: >>>> Probably? : >>>> >>>> All functions for which a link function of the response is modeled as a linear predictor of the covariates, but the response need not be in the exponential family? Such a list of course cannot be "listed". I would expect the package documentation, especially vignettes, explains this in any case. Pls check. >>>> >>>> Hopefully, you'll get a more authoritative response if I'm wrong. >>>> >>>> Bert >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> On May 11, 2017 4:15 AM, "Bruce Ratner PhD" <br at dmstat1.com <mailto:br at dmstat1.com>> wrote: >>>> >>>> R-helpers: >>>> In the "glmulti" package, it states parameter fitfunction assumes >>>> functions similar to glm, but doesn't list them. >>>> What are the functions similar to glm that can be used with glmulti? >>>> Bruce >>>> >>>> ______________________________________________ >>>> R-help at r-project.org <mailto:R-help at r-project.org> mailing list -- >>>> To UNSUBSCRIBE and more, see >>>> https://stat.ethz.ch/mailman/listinfo/r-help >>>> <https://stat.ethz.ch/mailman/listinfo/r-help> >>>> PLEASE do read the posting guide >>>> http://www.R-project.org/posting-guide.html >>>> <http://www.R-project.org/posting-guide.html> >>>> and provide commented, minimal, self-contained, reproducible code. >>>> >>>> >>> > > [[alternative HTML version deleted]] > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > 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.David Winsemius Alameda, CA, USA
Dear David: Thank you for your excellent reply. Apparently, you are a in the know. Again, thanks. Bruce ______________ Bruce Ratner PhD The Significant Statistician? (516) 791-3544 Statistical Predictive Analytics -- www.DMSTAT1.com Machine-Learning Data Mining -- www.GenIQ.net> On May 11, 2017, at 10:53 AM, David Winsemius <dwinsemius at comcast.net> wrote: > > >> On May 11, 2017, at 6:31 AM, Bruce Ratner PhD <br at dmstat1.com> wrote: >> >> Bert: >> Not clear to me. >> Where mentioned are the functions similar to glm, if you please? > > The basis for the similarity was stated as having an available link function (and I suspected, an inverse as well.) I, for one, wouldn't have been surprised if `lm` were not in the list because `glm` with a 'gaussian' link would provide the same capabilities. I read the basis as stating that a family-object be available (an object with the features described on the `?family` page. If I were correct, then running `methods(family)` would provide a list of the family objects that are available for loaded packages: > >> methods("family") > [1] family.glm* family.lm* family.negbin* > see '?methods' for accessing help and source code > > So lm would qualify as well. > >> require(lme4) > Loading required package: lme4 > Loading required package: Matrix >> methods("family") > [1] family.glm* family.glmResp* family.lm* family.lmResp* family.merMod* > [6] family.negbin* family.nlsResp* > see '?methods' for accessing help and source code > > Following Bert's advice (and so reading the manual for you), I find that Vincent Calcagno has stated the criterion somewhat differently in "Using glmulti with any type of statistical model, with examples." > #-------- > glmulti works outof-the-box with several types of function (such as lm, glm or coxph), but it can in principle be used with any such function `myttingfunction`, as long as> 1. The function receives a model specication in the form of a formula;> 2. The functionts the model by maximum likelihood, which can be accessed through the standard `LogLik` function;> > Even when the two conditions above are veried, complications arise because, unfortunately, dierent tting functions have dierent conventions regarding how characteristics of the t should be accessed. Indeed, most of them come from dierent packages with dierent authors and there is no common standard so far.> #-------- > > > > -- > David >> Bruce >> >> ______________ >> >> >> >> >>> On May 11, 2017, at 8:39 AM, Bert Gunter <bgunter.4567 at gmail.com> wrote: >>> >>> ?glmulti seems clear enough to me. If not, pls check the reference given therein. >>> >>> Bert >>> >>>> On May 11, 2017 5:22 AM, "BR_email" <br at dmstat1.com> wrote: >>>> Thanks, Bert. I would expect the list to include, at least lm. The reference states, "See Examples section." >>>> But, there is nothing in that section or elsewhere!! >>>> Bruce >>>> >>>> Bert Gunter wrote: >>>>> Probably? : >>>>> >>>>> All functions for which a link function of the response is modeled as a linear predictor of the covariates, but the response need not be in the exponential family? Such a list of course cannot be "listed". I would expect the package documentation, especially vignettes, explains this in any case. Pls check. >>>>> >>>>> Hopefully, you'll get a more authoritative response if I'm wrong. >>>>> >>>>> Bert >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> On May 11, 2017 4:15 AM, "Bruce Ratner PhD" <br at dmstat1.com <mailto:br at dmstat1.com>> wrote: >>>>> >>>>> R-helpers: >>>>> In the "glmulti" package, it states parameter fitfunction assumes >>>>> functions similar to glm, but doesn't list them. >>>>> What are the functions similar to glm that can be used with glmulti? >>>>> Bruce >>>>> >>>>> ______________________________________________ >>>>> R-help at r-project.org <mailto:R-help at r-project.org> mailing list -- >>>>> To UNSUBSCRIBE and more, see >>>>> https://stat.ethz.ch/mailman/listinfo/r-help >>>>> <https://stat.ethz.ch/mailman/listinfo/r-help> >>>>> PLEASE do read the posting guide >>>>> http://www.R-project.org/posting-guide.html >>>>> <http://www.R-project.org/posting-guide.html> >>>>> and provide commented, minimal, self-contained, reproducible code. >>>>> >>>>> >>>> >> >> [[alternative HTML version deleted]] >> >> ______________________________________________ >> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see >> 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. > > David Winsemius > Alameda, CA, USA > >