[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

# Re: st: vector of independent variables in ML program

 From "Stas Kolenikov" <[email protected]> To [email protected] Subject Re: st: vector of independent variables in ML program Date Tue, 12 Aug 2008 09:45:23 -0500

```Are you writing an ado estimation file, or just using -ml-
interactively? With your own programming, you can do whatever... I
would probably have a crude -global- to transfer whatever I need
between my components. With interactive use of -ml-, as I said,
theoretically you don't have to be needing that, at least for the
likelihoods with linear indices.

Again, with proper programming, the equation names in the system will
also be supplied to you -- I don't have my [ML] book with you to
refer, but I know for sure that's doable.

2008/8/12 Jonathan Hanson <[email protected]>:
> Dear Stas,
>
> Thanks for the reply.  I have an earlier edition of the Gould et. al. book
> and agree that it is helpful.  Nothing in there, however, pertaining to my
> question, at least in the edition I have.
>
> I would definitely rather not write the gradient vector, but my model is
> complex and the lf method of estimation is not working for me.  As I work
> through the derivatives, it seems that I have vectors of independent
> variables that need to be called directly, hence my question.
>
> Thanks again,
>
> Jon
>
> On Aug 12, 2008, at 9:47 AM, Stas Kolenikov wrote:
>
>> A big idea of -ml- is that you don't have to do that, as the
>> derivatives with respect to beta's necessary for likelihood
>> maximization and variance estimation will be computed by the chain
>> rule. If all that your equations depend on is theta, then you don't
>> really need to go into x's. The book on [ML] will explain all the
>> details (http://www.stata.com/bookstore/mle.html), and if you write
>> your own -ml- code, that book is a must.
>>
>> On Tue, Aug 12, 2008 at 8:29 AM, Jonathan Hanson
>> <[email protected]> wrote:
>>>
>>> Greetings,
>>>
>>> How does one refer to a vector of independent variables from within an ML
>>> program?  For example, suppose I need to write a gradient vector, the
>>> formula for which looks something like this:
>>>
>>> ∂ ln L/ ∂ Β =  (1/sigma^2) x'(y - xΒ)
>>>
>>> where x' is a k by 1 vector of the independent variables for an
>>> observation.
>>> I know that I can refer to y as \$ML_y1 and xB as one of the arguments
>>> (theta, for instance), but how can I call x' in the formula?  Ultimately,
>>> I
>>> will have multiple equations in this formula, so I need to be able to
>>> specify equation from which the x's are coming.
>>>
>>> Many thanks,
>>>
>>>
>>>
>>> Jonathan Hanson
>>> Assistant Professor of Political Science
>>> Maxwell School of Citizenship and Public Affairs
>>> 100 Eggers Hall
>>> Syracuse University
>>> Syracuse, NY 13244
>>>
>>>
>>>
>>>
>>>
>>> *
>>> *   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/
>>>
>>
>>
>>
>> --
>> Stas Kolenikov, also found at http://stas.kolenikov.name
>> Small print: I use this email account for mailing lists only.
>> *
>> *   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/
>
> _______________________________________
> Jonathan Hanson
> Assistant Professor of Political Science
> Maxwell School of Citizenship and Public Affairs
> 100 Eggers Hall
> Syracuse University
> Syracuse, NY 13244
> [email protected]
>
>
>
>
> *
> *   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/
>

--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   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/
```

 © Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index