Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# st: Margeff for lagged dependent variable

 From Stephen O Neill To statalist@hsphsun2.harvard.edu Subject st: Margeff for lagged dependent variable Date Wed, 24 Mar 2010 09:47:16 -0700 (PDT)

```Dear Statalist,
I am estimating the average partial effects for dynamic probit models by recaling the betas by 1/sqrt(1+(sigma_u)^2) and then using margeff on Stata 10. I have also calculated the partial effects using the following code:

scale= 1/sqrt(1+(sigma_U)^2)

* For binary variables:

gen xdh0_`i'=(xdh*scale-_b[`i']*`i'*scale)
gen xdh1_`i'=(xdh*scale-_b[`i']*`i'*scale+_b[`i']*scale)
gen double pe_`i' =normal(xdh1_`i') - normal(xdh0_`i')

sum pe_`i'

* For continuous variables:
gen double pe_`i' =normalden(xdh0*scale)*_b[`i']*scale
sum pe_`i'

When I compare the estimates they all match with the exception of the lagged dependent variable [from margeff I get 0.244475, while manually I calculate it as 0.409561. (I want to use Margeff to calculate the standard errors) Since I am using a loop to carry out the calculations I think that it is not just a mistake in my code. I have renamed the variables and also changed the order so I think I can rule out these explanations also.

I was wondering if anyone here has experienced similar issues using margeff on lagged dependent variables or if they can suggest a possible cause/solution. I don't have access to Stata11 so can't use the margins command :(

Thanks,
Stephen

*
*   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/
```