Or maybe you want:
sysuse nlsw88, clear
g lnw=ln(wage)
reg lnw c.ttl_exp##c.ttl_exp
margins, dydx(ttl_exp)
averaging across the sample. Or did you mean you want to treat X as estimated?
On Sun, Sep 27, 2009 at 11:48 AM, Maarten buis <[email protected]> wrote:
> --- Vatsalya Srivastava wrote:
>> I have a regression equation
>> y = bo + b1.x1 + b2.x2 + b3.x2^2 [simple OLS regression]
>>
>> I need to estimate the standard error of [b2 + b3.x2].
>>
>> The closest I have come is the nlcom command, which I suppose
>> estimates the variance of non linear combinations of estimators only.
>
> What you want is not a single value with a single standard error, but
> multiple values, one for each value of x2, with multiple standard errors.
> The relevant command in that case is -predictnl-, see -help predictnl-
> and the example below:
>
> *------------ begin example ------------------
> sysuse nlsw88, clear
> gen ttl_exp2 = ttl_exp^2
> gen ln_wage = ln(wage)
>
> reg ln_wage union grade ttl_exp ttl_exp2
>
> predictnl effect = ///
> _b[ttl_exp] + 2*_b[ttl_exp2]*ttl_exp, ///
> ci(lb ub)
>
> sort ttl_exp
> twoway rarea lb ub ttl_exp || ///
> line effect ttl_exp, ///
> clstyle(solid) legend(off) ///
> ytitle(effect of ttl_exp) ///
> yline(0)
> *----------- end example -------------------
> (For more on examples I sent to the Statalist see:
> http://www.maartenbuis.nl/example_faq )
>
> Hope this helps,
> Maarten
*
* 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/