[R] [FORGED] country fixed effects model with binary dependent variable

[Rolf Turner]

On 24/07/16 02:49, Cristina Cametti wrote:

> Dear all,
>
> I am having problems finding a reliable code for a country fixed
effects model with binary dependent variable. I was able to run it for
another part of my research, because in that case the dependent variable
is continuous.
> This is my code for the continuous dependent variable “imwbcrm_rec”:
>
> modelfe2 <- lm(imwbcrm_rec ~ tvpol + victim + agea + gndr + eduyrs +
lrscale_GM + imgfrnd_dum + qfimwht +factor(cntry)-1, data=mydata)
>
> Please don’t mind to how I wrote the variables, they are from the
first wave of the ESS survey. At this point, I have three questions:
> - do you think this code is correct?

It is impossible to say; members of this list are not in general telepathic.

You need to describe your problem and your data more clearly, or 
possibly supply a reproducible example, in order for anyone to have a 
ghost of a chance of being able to advise you.

> Since the data are all from the
same year (2002), I did not used the ppm package since it is only for
panel data. The results of the previous code make sense, so I am
satisfied. However, I want to be sure that I am running the right code.
> - second questions: someone knows the code for the same analysis,
> but
having a BINARY dependent variable (aesfdrk_dummy)? I found very
different information on the internet, and unfortunately, I do not know
how to use STATA, so I need to find a reliable code in r. This is the
code that I have now:
>
> modelfe1 <-lm(aesfdrk_dummy ~mixed_neigh + ethnic_neigh + agea+ gndr
> +
eduyrs + domicil + partner + tvpol + hincfel_dum + factor(cntry) -1,
data=mydata)

It is very common to model a binary response using a logit model
(i.e. use glm() with family=binomial).  Whether this is appropriate in 
the given case is impossible to say.

> -last question: I have to add some interaction between country level
variables and individual level variables. So, do you think that this
code is right?
>
> mydata$ppltrst_GMXprison_pop <- mydata$ppltrst_GM*mydata$prison_pop
> mydata$ppltrst_GMXforeign_pop <-
> mydata$ppltrst_GM*mydata$foreign_pop mydata$victimXprison_pop<-
> mydata$victim*mydata$prison_pop mydata$victimXforeign_pop<-
> mydata$victim*mydata$foreign_pop mydata$mixed_neighXprison_pop<-
> mydata$mixed_neigh*mydata$prison_pop mydata$mixed_neighXforeign_pop
> <- mydata$mixed_neigh*mydata$foreign_pop
> mydata$ethnic_neighXprison_pop <-
> mydata$ethnic_neigh*mydata$prison_pop mydata$ethnic_neighXforeign_pop
> <- mydata$ethnic_neigh*mydata$foreign_pop
>
> modelfe1.2<-lm(aesfdrk_dummy ~mixed_neigh + ethnic_neigh + agea +
> gndr
+ eduyrs + domicil + partner + tvpol + hincfel_dum + victim + ppltrst +
ppltrstXprison_pop + ppltrst_Xforeign_pop_ + victimXprison_pop_ +
victimXforeign_pop + mixed_neighXprison_pop + mixed_neighXforeign_pop +
ethnic_neighXprison_pop + ethnic_neighXforeign_pop + factor(cntry)-1,
data=mydata)

The foregoing makes no sense at all to me.  Perhaps I am just being 
obtuse, but I doubt it.

You seem to be very much out of your depth.  I think you would be 
well-advised to seek local statistical consultation.

cheers,

Rolf Turner

-- 
Technical Editor ANZJS
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276