Re: st: Constructing a variable from standard deviations

[Austin Nichols <austinnichols@gmail.com>]

M.P.J. van Zaal <Matthias.vanZaal@student.uva.nl>:

egen ... , by(occ)
not
egen ... , by(each residual value)


On Mon, Nov 22, 2010 at 4:44 PM, M.P.J. van Zaal
<Matthias.vanZaal@student.uva.nl> wrote:
> I have now run all the regression for the occupation
>
> reg ink controls dummy1, robust
> predict resid1, resid
> .
> reg ink controls dummy106, robust
> predict resid106, resid
>
>
> Then I used:
> egen s = sd( res ), by(resid1-resid106)
> g newerror = s*rnormal()
>
> But now i get a variable with value "0" and frequency "6".
>
> What did i do wrong?
>
> I checked the stand deviations with tabstat and this works. They are all
> varying even more now.
> ----- Original Message -----
> From: Austin Nichols <austinnichols@gmail.com>
> Date: Monday, November 22, 2010 9:23 pm
> Subject: Re: st: Constructing a variable from standard deviations
> To: statalist@hsphsun2.harvard.edu
>
>> M.P.J. van Zaal <Matthias.vanZaal@student.uva.nl>:
>>
>> Don't say "you guys claim that the estimates from this procedure would
>> be meaningless" --Stas agreed with the basic approach, as do I, since
>> it is standard and easily understood.  Stas almost gave you the
>> answer:
>>
>> regress depvar whatever i.occupation
>> predict res, res
>> egen s = sd( res ), by( occupation )
>> g newerror = s*rnormal()
>>
>>
>> On Mon, Nov 22, 2010 at 3:09 PM, M.P.J. van Zaal
>> <Matthias.vanZaal@student.uva.nl> wrote:
>> > I am still digesting all this new information..
>> >
>> > The approach used by Kit seems to be familiar to an approach i used
>> > earlier. My prof gave me the advice to include all the dummies in
>> the> regression and then calculate the residual per occupation,
>> since this is
>> > what Bonin et all do.
>> >
>> > However, you guys claim that the estimates from this procedure
>> would be
>> > meaningless. I will start over again and make regression per
>> occupation> and then predict the residuals.
>> >
>> > When I estimate the standard error of all these residuals I will
>> > probably have 106 different "variables". How can I construct one
>> > variable from all these standard deviations? This is the problem at
>> > which I was stuck at my old approach, so i will probably have the
>> > problem again when i use a new method.
>

*
*   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/