Thread Subject:

Binary logit regression

Good afternoon,

I am attempting to model sovereign default probabilities. To do so I have as dependant variable:
y=[0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1]
and four explanatory variables x1 x2 x3 and x4, all of same dimension.

Here are my issues:
1) Shall I regress first my explanatory variables against the variable Y (which is not in a binary form) from which I derived y? Or should I directly regress these explanatory variables against y?

If the second option is the right one, how do I do that?

2) I do not really know what are the steps I should follow then to obtain the probability of default, according to the variable that I will find significant.

Assuming x1 and x2 are significant, are the following steps correct?

X=[x1 x2]
[b,dev,stats] = glmfit(X,y,'binomial','link','logit')
yfit = glmval(b, X,'logit');

3) Finally, if the above steps are correct, among the outputs returned, which one would give me the probability of default?

Thanking you in advance for your help.

Victor

On Jun 30, 1:10 pm, "Victor " <victorgen...@hotmail.fr> wrote:
> Good afternoon,
>
> I am attempting to model sovereign default probabilities. To do so I have as dependant variable:
> y=[0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1]
> and four explanatory variables x1 x2 x3 and x4, all of same dimension.
>
> Here are my issues:
> 1) Shall I regress first my explanatory variables against the variable Y (which is not in a binary form) from which I derived y? Or should I directly regress these explanatory variables against y?
>
> If the second option is the right one, how do I do that?
>
> 2) I do not really know what are the steps I should follow then to obtain the probability of default, according to the variable that I will find significant.
>
> Assuming x1 and x2 are significant, are the following steps correct?
>
> X=[x1 x2]
> [b,dev,stats] = glmfit(X,y,'binomial','link','logit')
> yfit = glmval(b, X,'logit');
>
> 3) Finally, if the above steps are correct, among the outputs returned, which one would give me the probability of default?

yfit.

help glmfit
doc glmfit

Hope this helps.

Greg

Hi Greg,

Thanks for your prompt answer. I unfortunately already read the matlab doc as well as other posts in matlab central but I am still not sure of what I have to do. If you could give me some advices regarding the above mentioned three steps, I would really appreciate.

Thanks again for your help.

Victor