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 at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: nl: missing standar error


From   chw_wagner@gmx.de
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: nl: missing standar error
Date   Tue, 18 Sep 2012 09:54:54 +0200

Hi,

thanks a lot for the hint.

My code looks like this (in a abbreviated form):

scalar rate=0.05

nl ( variable1 =
( {constant=1}
+ ({alpha1=1}*rate+{alpha2=1}*log(exp({alpha3}))*rate)*variable2
+ {alpha2=1}*rate*variable3
+ ({beta}*rate+{alpha4}*log(exp({alpha3}))*rate)*variable4 
+ {alpha4=1}*rate*0.5*variable5
- {alpha4=1}*0.5*variable6)
- exp({alpha3})*variable7), nolog vce(robust);

The missing standard error will be produced for alpha1, but just If I use my real data set. There will be no missing standard error if I use my test case.

That do you mean by "outputting the covariance matrix of the
parameter estimates"?

Thanks a lot for your help,
Christina

> On Thu, Sep 13, 2012 at 12:06 PM,  <chw_wagner@gmx.de> wrote:
> 
> >
> > Could you please give me a hint how to identify the problem? I have no
> idea, why Stata is not reporting this standard error when I am using my real
> data set. For me, the programming code looks the same except that I have
> to replace my variables according to the variables in the test case.
> >
> 
> It sounds to me like the variance of that parameter is essentially 0
> or some other violation of the requirement that a covariance matrix be
> positive definite. Try outputting the covariance matrix of the
> parameter estimates and see if there are other missing parts or if
> (when looked at in correlation scale) the correlation between one
> parameter and another is very high.
> 
> You may have an empirically unidentified parameter.
> 
> *
> *   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/

---original massage---

>Dear StataList,

>I have the following problem. I am trying to run a non-linear regression >in Stata 11. I am using a test case to try my programmed codes before I >use my real data set. This test case does not contain the variables, >which I will use later on in my analyses, but at least it has a panel >structure which is important for me.

>I programmed my function using the nl command. Then I run the model with >the test case. Until this point the codes are working. If I run the >regression commands with my real data set, Stata will not display one >standard error for one specific coefficient of a variable, even though >Stata did it for my test data. I have tried to make my observations more >homogeneous, but it is still not working.

>Could you please give me a hint how to identify the problem? I have no >idea, why Stata is not reporting this standard error when I am using my >real data set. For me, the programming code looks the same except that I >have to replace my variables according to the variables in the test case.

>Thank you in advance

>Christina
*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index