Re: st: comparing the effect of independent variables at two points in time, OLS and probit

[Nick Cox <njcoxstata@gmail.com>]

The answers to, if not the questions starting, various concurrent
threads touch on relevant material. Please look at them. Those started
by David Ashcraft and Muhammad Anees are especially relevant.

Nick

On Wed, Dec 14, 2011 at 9:45 AM, Laura R. <laura.roh@googlemail.com> wrote:

> I would like to see whether the influence of predictors for a certain
> variable are different in 2 points in time. E.g., does the work status
> (not employed, part-time, full-time) have a different effect  on y in
> 1980 than in 2000?
>
> I am not sure if I can just make 2 different regressions, one for 1980
> and one for 2000, and then compare the coefficients. Is this correct
> for OLS and for probit (marginal effects)? At least for marginal
> effects, I think you cannot compare coefficients between 2 models.
>
> So what other possibilities are there? Maybe pool data from 1980 and
> 2000 and make interaction effects. But I am not sure what categories I
> need and which would be the reference category/categories. So I would
> have the following categories: no employment 1980, part-time
> employment 1980, full-time employment 1980, no employment 2000,
> part-time 2000, full-time 2000. Do I also need the employment
> categories overall, i.e. without interaction? And which would be the
> reference categories, if I would like to compare the impact of
> full-time work in 1980 with full-time work in 2000, part-time 1980
> with part-time 2000 and so on? I am confused.
>
> Do you have any other idea how to compare the effect of certain
> predictors between 2 points in time? I would also do the search
> myself, but I am also not sure which keywords to use for searching
> about this question, or in which textbook I could find something about
> this.
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/