R help - Aug 2016 - GLM output problem

Hello

I have started to work with GLM and I am facing the following problem:

If I take:

y = c(0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0)
x = 1:18

model = y ~x + I(x^2)
GLM = glm(model, family=binamial(link = logit))

And use the parameters returned by GLM to contruct an equation for the
regression model:

model.eq = -0.446078 + 0.267673*x - 0.014577*I(x^2)

And backtransform it from the logit to the natural scale (using the inverse
link-function for this case):

model.proj = exp(model.eq)/(1+exp(model.eq))

the plot for model.proj~x is not the same of the plot for fitted(GLM)~x
(see the output attached).

Why is this happening? Can someone help me?

Regards,

Anderson Eduardo

 --
Anderson A. Eduardo
------------------------------------------------------------------------------
Lattes <http://lattes.cnpq.br/3826166230581311> | Researcher ID
<http://orcid.org/0000-0001-8045-8043> | Google Acad?mico
<https://scholar.google.com.br/citations?user=oOUjq9IAAAAJ&hl=pt-BR> |
Site
<http://andersonaireseduardo.xpg.uol.com.br/>
------------------------------------------------------------------------------

Inline.

-- Bert
Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Wed, Aug 31, 2016 at 10:03 AM, Anderson Eduardo
<andersonaed at gmail.com> wrote:
> Hello
>
> I have started to work with GLM and I am facing the following problem:
>
> If I take:
>
> y = c(0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0)
> x = 1:18
>
> model = y ~x + I(x^2)
> GLM = glm(model, family=binamial(link = logit))
>
> And use the parameters returned by GLM to contruct an equation for the
> regression model:
>
> model.eq = -0.446078 + 0.267673*x - 0.014577*I(x^2)
## Not what I got with your data. I got:

Coefficients:
(Intercept)            x       I(x^2)
   -18.5750       5.0403      -0.2845


I suspect you had some other x,y variables lying around when you
defined your model.

-- Bert
>
> And backtransform it from the logit to the natural scale (using the inverse
> link-function for this case):
>
> model.proj = exp(model.eq)/(1+exp(model.eq))
>
> the plot for model.proj~x is not the same of the plot for fitted(GLM)~x
> (see the output attached).
>
> Why is this happening? Can someone help me?
>
> Regards,
>
> Anderson Eduardo
>
>  --
> Anderson A. Eduardo
>
------------------------------------------------------------------------------
> Lattes <http://lattes.cnpq.br/3826166230581311> | Researcher ID
> <http://orcid.org/0000-0001-8045-8043> | Google Acad?mico
>
<https://scholar.google.com.br/citations?user=oOUjq9IAAAAJ&hl=pt-BR> |
Site
> <http://andersonaireseduardo.xpg.uol.com.br/>
>
------------------------------------------------------------------------------
> ______________________________________________
> 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.

>> And use the parameters returned by GLM to contruct an equation for the
>> regression model:
>> 
>> model.eq = -0.446078 + 0.267673*x - 0.014577*I(x^2)
> 
> ## Not what I got with your data. I got:
> 
> Coefficients:
> (Intercept)            x       I(x^2)
>   -18.5750       5.0403      -0.2845
> 
> 
> I suspect you had some other x,y variables lying around when you
> defined your model.

More likely, the family= specification got lost and gaussian family implied:
> glm(model)
Call:  glm(formula = model)

Coefficients:
(Intercept)            x       I(x^2)  
   -0.44608      0.26767     -0.01458  

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com