Dear Lister When the count outcomes are integers, we could use either Poisson or NB regression to model them. However, there are cases that the count outcomes are non-integers, e.g. average counts. I am wondering if it still makes sense to use Poisson or NB regression to model these non-integer outcomes. Truly appreciate your attention and insight!
On Jul 21, 2015, at 8:21 PM, Wensui Liu wrote:> Dear Lister > When the count outcomes are integers, we could use either Poisson or > NB regression to model them. However, there are cases that the count > outcomes are non-integers, e.g. average counts. > I am wondering if it still makes sense to use Poisson or NB regression > to model these non-integer outcomes.There is a quasi-binomial error model that accepts non-integer outcomes. -- David Winsemius Alameda, CA, USA
Or if there are enough averages of enough counts, the CLT provides another option.> On Jul 21, 2015, at 8:38 PM, David Winsemius <dwinsemius at comcast.net> wrote: > > > On Jul 21, 2015, at 8:21 PM, Wensui Liu wrote: > >> Dear Lister >> When the count outcomes are integers, we could use either Poisson or >> NB regression to model them. However, there are cases that the count >> outcomes are non-integers, e.g. average counts. >> I am wondering if it still makes sense to use Poisson or NB regression >> to model these non-integer outcomes. > > There is a quasi-binomial error model that accepts non-integer outcomes. > > -- > > David Winsemius > Alameda, CA, USA > > ______________________________________________ > 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.
Sorry. Central limit theorem. Enough averaging and you get a normal distribution (simply stated, perhaps too simply). If so others will correct me before long. :-( Sent from my iPad> On Jul 21, 2015, at 8:52 PM, Wensui Liu <liuwensui at gmail.com> wrote: > > what does CLT stand for? > >> On Tue, Jul 21, 2015 at 11:41 PM, Don McKenzie <dmck at u.washington.edu> wrote: >> Or if there are enough averages of enough counts, the CLT provides another option. >> >>> On Jul 21, 2015, at 8:38 PM, David Winsemius <dwinsemius at comcast.net> wrote: >>> >>> >>> On Jul 21, 2015, at 8:21 PM, Wensui Liu wrote: >>> >>>> Dear Lister >>>> When the count outcomes are integers, we could use either Poisson or >>>> NB regression to model them. However, there are cases that the count >>>> outcomes are non-integers, e.g. average counts. >>>> I am wondering if it still makes sense to use Poisson or NB regression >>>> to model these non-integer outcomes. >>> >>> There is a quasi-binomial error model that accepts non-integer outcomes. >>> >>> -- >>> >>> David Winsemius >>> Alameda, CA, USA >>> >>> ______________________________________________ >>> 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. >> >> >> > > > > -- > WenSui Liu > https://statcompute.wordpress.com/
> On 22 Jul 2015, at 06:48 , Don McKenzie <dmck at uw.edu> wrote: > > Sorry. Central limit theorem.Or some sort of vegetarian sandwich. Celery, Lettuce, Tomato sounds almost edible with sufficient mayo. ;-)> Enough averaging and you get a normal distribution (simply stated, perhaps too simply). If so others will correct me before long. :-(Well, your punctuation doesn't quite work -- ')' comes too early. Otherwise it is close enough for jazz, although there are distributions that you can average forever and still not get a normal, and some might want to stress that it is the parameter estimators that become approximately normal. (Students sometimes get confused and believe that the original data magically become normally distributed when you have a lot of them.) In practice, one should ensure that one has "many" data for all the relevant averages (996 males and 4 females is no good), and also that one gets the variance structure at least roughly right. -pd> > Sent from my iPad > >> On Jul 21, 2015, at 8:52 PM, Wensui Liu <liuwensui at gmail.com> wrote: >> >> what does CLT stand for? >> >>> On Tue, Jul 21, 2015 at 11:41 PM, Don McKenzie <dmck at u.washington.edu> wrote: >>> Or if there are enough averages of enough counts, the CLT provides another option. >>> >>>> On Jul 21, 2015, at 8:38 PM, David Winsemius <dwinsemius at comcast.net> wrote: >>>> >>>> >>>> On Jul 21, 2015, at 8:21 PM, Wensui Liu wrote: >>>> >>>>> Dear Lister >>>>> When the count outcomes are integers, we could use either Poisson or >>>>> NB regression to model them. However, there are cases that the count >>>>> outcomes are non-integers, e.g. average counts. >>>>> I am wondering if it still makes sense to use Poisson or NB regression >>>>> to model these non-integer outcomes. >>>> >>>> There is a quasi-binomial error model that accepts non-integer outcomes. >>>> >>>> -- >>>> >>>> David Winsemius >>>> Alameda, CA, USA >>>> >>>> ______________________________________________ >>>> 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. >>> >>> >>> >> >> >> >> -- >> WenSui Liu >> https://statcompute.wordpress.com/ > ______________________________________________ > 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.-- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd.mes at cbs.dk Priv: PDalgd at gmail.com