# Re: st: Decile regression using sqreg

 From "Leonor Saravia" <[email protected]> To [email protected] Subject Re: st: Decile regression using sqreg Date Fri, 25 Apr 2008 10:56:16 -0400

```Ronand and Maarten,

Thank you very much for your responses, now I get why Stata is
considering the quantiles between 0 and 1, exclusive, when we use the
"sqreg" command.
But, may I bother you again? I understood that I'm having a problem
with the interpretation of the coefficients, because I can't
understand what's the meaning of each one of the coefficients obtained
with this command.

I'm working with some households surveys for latinamerican countries,
so the data I'm using has the next variables:

household_identifier     income    food_expenditure
123                               1000         800
445                                2980        1500
670                               500            250
.                                     .                 .
.                                     .                 .
.                                     .                 .
870                              15200          14000

The model that I'd like to perform is this:

income = a + b*(food_expenditure) + error

And I need to compare the coefficients for each decil of income of the
population of each country, something like comparing the coefficient
for the food expenditure done by 10% of the poorest population and the
richest 10% of the population.

I think that the quantile regression is the way to do this, using the
"sqreg" command, am I wrong?

That's why I thought that I need 10 regressions instead of 9, because
I need to compare the coefficients for each one of the 10th parts of
the population.

Thank you very much!

Leonor

2008/4/24, Maarten buis <[email protected]>:
> > --- Leonor Saravia <[email protected]> wrote:
> > > I need to estimate a regression for each decil of income for a
> > > country, so, if I�m understanding well, using "sqreg" it�s not the
> > > way to do it, isn�t it? Because I�m obteining 9 regression instead
> > > of 10, being 10 what I was expecting.
>
> --- Maarten buis <[email protected]> wrote:
> > You should expect 9 regressions: you are modeling the first to the
> > ninth decile and never the zeroth or the tenth decile (being the
> > minimum and the maximum). This is implicit in the description of the
> > description of the -quantile()- option in the helpfile of -sqreg-:
> > "quantile(#) specifies the quantile to be estimated and should be a
> > number between 0 and 1, exclusive."
>
> Let me expand a bit on that: The key in this quote is the word
> "exclusive": the quantiles 0 and 1 are not allowed. This makes sense.
> Consider your case, where you want deciles: This means you want to cut
> you dependent variable into ten pieces, and the deciles are the points
> where you make the cuts. If you want to cut something in ten pieces you
> only need nine cuts; you don't need to cut at the bottom (zeroth
> decile) or the top (tenth decile).
>
> Hope this helps,
> Maarten
>
> -----------------------------------------
> Maarten L. Buis
> Department of Social Research Methodology
> Vrije Universiteit Amsterdam
> Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
>
> Buitenveldertselaan 3 (Metropolitan), room Z434
>
> +31 20 5986715
>
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
>
>
>      ___________________________________________________________
> Yahoo! For Good. Give and get cool things for free, reduce waste and help our planet. Plus find hidden Yahoo! treasure
>
> http://green.yahoo.com/uk/earth-day/
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```