Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Fwd: How to interpret regression?

From   Joerg Luedicke <>
Subject   Re: st: Fwd: How to interpret regression?
Date   Tue, 25 Sep 2012 08:57:08 -0500

It seems that you are more concerned about p-values than you are
concerned about what you can actually learn about your data by fitting
this model. It looks to me as if you are interested in measuring a
non-linear effect of wealth on your outcome that differs across the
groups defined by your dummy variables. If this is correct I would
start with inspecting the bivariate relations between your outcome and
wealth and check if this relation differs across your groups (for
example using a smoother like -lowess- or -lpoly-). Depending on how
these relations look like it could make sense to summarize them using
a model that includes (perhaps 2nd or 3rd order) polynomial terms for
wealth, which would be one convenient possibility. Of course, it can
always be checked how good of a summary that would be, given your
data. In any case, I find it hard to see why you would include wealth
as a continuous variable and a categorical version of wealth in the
same model.

In any case, you should perhaps provide more information about your
research question, variables etc. Just saying that some inputs xyz are
significant in one model but not in another is probably not very
helpful in order to get reasonable advice.


On Tue, Sep 25, 2012 at 7:05 AM, Kresten Buch <> wrote:
> Dear Statalist,
> I have problems figuring out what I can interpret from this
> regression. Could you please help me?
> I regress a taxation measure against 4 dummies and their interaction
> with a measure of wealth, and interaction between the 4 dummies and
> a dummy constructed from the measure of wealth, where I give the value
> 1 to observations in the top 10% or bottom 10% of the wealth measure,
> and 0 to the rest of the observations.
> If I include all variables in the model, I get significant effects
> from both sets of interaction variables. If only one set is included,
> then I nothing is significant, no matter witch one is excluded.
> I hope the explanation makes sense ?
> --
> Best regards from Kresten Buch, Denmark
> *
> *   For searches and help try:
> *
> *
> *
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index