Re: Interpreting the coefficient in a GLS binomial model

Serguei Kaniovski <kaniovsk@xxxxxxxxxx> wrote in
news:13646874.1172339189307.JavaMail.jakarta@xxxxxxxxxxxxxxxxxxxxxx:

Hi,

I would appreciate you help. In a GLS binomial model. the proportion
of concurrent votes in total votes between a pair of voters is a
function of their ideological distance (index continuous on [0,1]).I
get the following estimates: Intercept: 2.982868 Coef: -1.450829.

How do I compute the marginal effect of an increase in the index by
0.1?

I transform from logits to proportios as:
p=1/(1+1/exp(-1.450829))=0.19

But is not yet what I need. Can you help?


I am assuming you are asking about what is generally called logistic
regression? Link is logistic, errors are binomial?

First point: the predicted proportions calculation should include all the
parameters, including the intercept. I have no idea what you estimated in
your calculation above. If you want the estimate proprtion of votes for
someone with a ideologic distance of 0.5 the expression would be;
1/(1+1/exp(2.983 + (-1.450829*0.5)) (1)

When you say the "marginal effect", would you accept the odds ratio? If so,
then the OR for a positive change in ideologic distance of 0.1 is just
exp(-1.451*0.1). If you need something different, say a ratio of
proportions, you can construct such with expressions similar to (1). The
ratio (or difference) of proportions would vary for different locations
along the range of "ideologic distance". It is only linear for OR as an
effect measure.

--
Davuid Winsemius
.