[R] Struggling to compute marginal effects !

[Willy Byamungu]

Thank you very much Jim.

On Mon, Mar 19, 2018, 7:20 PM Jim Lemon <drjimlemon at gmail.com> wrote:

> In that case, I can't work out why the first model fails but not the
> second. I would start looking at "Data" to see what it contains. if:
>
> object2 <- polr(Inc ~ Training ,Data,Hess = T,method = "logistic" )
>
> works, the problem may be with the "Adopt" variable.
>
> Jim
>
>
> On Tue, Mar 20, 2018 at 10:55 AM, Willy Byamungu
> <wmulimbi at email.uark.edu> wrote:
> > Sorry, do not read "Inc_reliable_fromNTA" I did a mistake in the email I
> > sent you. It should be just "Inc".
> >
> > The situation looks like this
> >
> > Data$Inc <- ordered(Data$Inc =
> > c("Not_reliable_at_all","Less_reliable","Somehow_reliable","Very
> reliable"))
> >
> > First model
> > object1 <- polr(Inc ~ Adopt ,Data,Hess = T,method = "logistic")
> >
> > Marginal effect for object1:
> > ocME(object1)
> > Error in eval(predvars, data, env) :
> >   numeric 'envir' arg not of length one
> >
> > Second model
> > object2 <- polr(Inc ~ Adopt + Training ,Data,Hess = T,method =
> "logistic" )
> >
> > Marginal effect for object2:
> > ocME(object2)
> > effect.Not_reliable_at_all effect.Less_reliable effect.Somehow_reliable
> > effect.Very reliable
> > Adopt                       -0.073               -0.057
> > -0.064                0.193
> > Training                    -0.283               -0.135
> > -0.051                0.469
> >
> >
> > On Mon, Mar 19, 2018 at 6:46 PM, Jim Lemon <drjimlemon at gmail.com> wrote:
> >>
> >> Hi Willy,
> >> I had a look at the "erer" manual, and it looks like a (somewhat)
> >> misleading error message. I don't know where "Inc_reliable_fromNTA"
> >> may be, but it looks like the ocME function can't find it, which leads
> >> to the "environment" error message. You have defined "Inc" as an
> >> element of "Data" and when that is used in the model, the ocME
> >> function works. I suspect that  "Inc_reliable_fromNTA" is not an
> >> element of "Data" and this is why ocME cannot find it.
> >>
> >> Jim
> >>
> >>
> >> On Tue, Mar 20, 2018 at 10:26 AM, Willy Byamungu
> >> <wmulimbi at email.uark.edu> wrote:
> >> > Dear Jim,
> >> >
> >> > Thank you very much for your help.
> >> >
> >> > My R-coding looks like this:
> >> >
> >> > Data$Inc <- ordered(Data$Inc =
> >> > c("Not_reliable_at_all","Less_reliable","Somehow_reliable","Very
> >> > reliable"))
> >> >
> >> > First model
> >> > object1 <- polr(Inc_reliable_fromNTA ~ Adopt ,Data,Hess = T,method =
> >> > "logistic")
> >> >
> >> > Marginal effect for object1:
> >> > ocME(object1)
> >> > Error in eval(predvars, data, env) :
> >> >   numeric 'envir' arg not of length one
> >> >
> >> > Second model
> >> > object2 <- polr(Inc ~ Adopt + Training ,Data,Hess = T,method =
> >> > "logistic" )
> >> >
> >> > Marginal effect for object2:
> >> > ocME(object2)
> >> > effect.Not_reliable_at_all effect.Less_reliable
> effect.Somehow_reliable
> >> > effect.Very reliable
> >> > Adopt                       -0.073               -0.057
> >> > -0.064                0.193
> >> > Training                    -0.283               -0.135
> >> > -0.051                0.469
> >> >>
> >> >
> >> > The model computing requires package "MASS" and marginal effects
> require
> >> > "ERER". As you can see above, in the first case it succeed to provide
> >> > results but in the first case NO.
> >> >
> >> > What do you think? By the way, what do you mean by inappropriate
> >> > environment
> >> > and predvars?
> >> >
> >> > Looking forward to hearing from you.
> >> >
> >> > Regards,
> >> >
> >> > Willy
> >> >
> >> > On Mon, Mar 19, 2018 at 5:33 PM, Jim Lemon <drjimlemon at gmail.com>
> wrote:
> >> >>
> >> >> Hi Willy,
> >> >> The error message may be due to passing an inappropriate environment.
> >> >> I don't have the "erer" package, but if that function has a default
> >> >> for the environment argument, perhaps passing just the "predvars" and
> >> >> "data" arguments will work.
> >> >>
> >> >> Jim
> >> >>
> >> >> On Tue, Mar 20, 2018 at 8:24 AM, Willy Byamungu
> >> >> <wmulimbi at email.uark.edu>
> >> >> wrote:
> >> >> > Dear Oscar,
> >> >> > and any other R-project person,
> >> >> >
> >> >> > Can you please help me to figure out the meaning of the following
> >> >> > error
> >> >> > message in red ?
> >> >> >
> >> >> > Error in eval(predvars, data, env) :
> >> >> >   numeric 'envir' arg not of length one
> >> >> >
> >> >> > I computed ordered logit models using 'polr' in R (I just followed
> >> >> > the
> >> >> > guidance a handout I found on princeton.edu about logit, probit
> and
> >> >> > multinomial logit models) . The summary results are obtained
> without
> >> >> > any
> >> >> > problem. *However*, when I'm using the package 'erer' to compute
> the
> >> >> > marginal effects for the ordered logit models, I'm just getting the
> >> >> > error
> >> >> > message above.
> >> >> >
> >> >> > Please HELP !!!!
> >> >> >
> >> >> > Willy
> >> >> >
> >> >> >
> >> >> > --
> >> >> > Willy Mulimbi B.
> >> >> > AEAB Grad Student & Fulbright Scholar
> >> >> > Tel. (+1) 479-316-5981
> >> >> > Fayetteville, AR
> >> >> >
> >> >> >         [[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.
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > Willy Mulimbi B.
> >> > AEAB Grad Student & Fulbright Scholar
> >> > Tel. (+1) 479-316-5981
> >> > Fayetteville, AR
> >
> >
> >
> >
> > --
> > Willy Mulimbi B.
> > AEAB Grad Student & Fulbright Scholar
> > Tel. (+1) 479-316-5981
> > Fayetteville, AR
>

	[[alternative HTML version deleted]]