# Re: st: xtmelogit calculation time

 From "Austin Nichols" To statalist@hsphsun2.harvard.edu Subject Re: st: xtmelogit calculation time Date Tue, 23 Sep 2008 15:43:59 -0400

```Emily Petherick <ep9@york.ac.uk> :
One way is to use the from() option (see help
xtmelogit##maximize_options and following discussion), like so:

webuse towerlondon, clear
set rmsg on
xi: xtmelogit dtlm difficulty i.group || family: || subject:
* above takes about 56 seconds
xi: xtmelogit dtlm difficulty i.group || subject:, laplace
mat b=e(b)
xi: xtmelogit dtlm difficulty i.group || family: || subject:, from(b)
* above takes about 55 seconds = 7.5 + 47.5

so you may see a small time savings, but it can make a big difference, too.

On Tue, Sep 23, 2008 at 5:37 AM, Emily Petherick <ep9@york.ac.uk> wrote:
> Hi Jeph,
>             really interested in your response as am having similar issues
> to Changwoo.  How do you go about using these earlier starting models as
> starting values for the later more complex models.
>
> Regards
> Emily Petherick.
>
> Jeph Herrin wrote:
>>
>> You don't mention the number of observations and panels, but these
>> are at least as critical as the number of covariates and random
>> coefficients. -xtmelogit- will take a very long time to estimate
>> 4 random coefficients if you do not have very many groups.
>>
>> You may want to start with 1 random coefficient, and use the results
>> of that model as starting values for a model with 2 random coeffs,
>> and so on.
>>
>> hth,
>> Jeph
>>
>>
>> Changwoo Lee wrote:
>>>
>>> Hi all,
>>> I'm using xtmelogit with 50 covariates including 4 random coefficients.
>>> Two days ago I ran it but I have only 'refining start values' on the screen.
>>> I don't know if Stata still work or not. If it still work, how much time
>>> should i wait for getting the results? Thanks in advance.
*
*   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/
```