Re: st: RE: Re: Marginal effects after ivreg2

[Suyin Chang <suyinlchang@gmail.com>]

Dear Mark,

This was quite helpful. I was trying something very similar, but
making a silly mistake in how to call the values generated in the
loop. Thanks for the lines of code.
While we wait to see if there will be any suggestions coming from the
list regarding how to plot those values, I would like to bring you
still another follow-up question to push forward these workarounds. As
we are here talking about how to make marginal effects available to
-ivreg2- while it does not integrate directly with -margins-, one
question comes to my mind: what about clustered se?

One of the main reasons I frequently to 2sls in Stata (besides trying
to isolate misspecification errors in individual equations when I run
nonrecursive models), is its much greater flexibility than the 3sls
estimation from -reg3- regarding the the standard errors, e.g. the
robust and cluster possibilities. Therefore, when we use cluster se in
the 2sls estimation through cluster(var) as an option, what would be
your suggestion on how to present meaningful marginal effects for
interaction terms?

Thanks for your insights.

Suyin

2014/1/14 Schaffer, Mark E <M.E.Schaffer@hw.ac.uk>:
> Suyin,
>
>> -----Original Message-----
>> From: Suyin Chang [mailto:suyinlchang@gmail.com]
>> Sent: 14 January 2014 06:19
>> To: statalist@hsphsun2.harvard.edu; Schaffer, Mark E
>> Subject: Re: st: RE: Re: Marginal effects after ivreg2
>>
>> Dears Jana and Mark,
>>
>> Very interesting question, followed by a promising reply.
>> However, I would like to ask two quick follow-up questions, if I may.
>>
>> The first problem is I don't get how to mimic -margins- with -nlcom- after
>> ivreg2 in case we want a linear plot. For instance, how to mimic something
>> like margins, dydx(mgp) at(turn=(start(by)end)), where start is the first value
>> and end is the last value of the variable turn to be considered? That appears
>> to me as something less intuitive to achieve.
>
> This is also easy to do - all you need to do is loop through the values and apply -nlcom- each time.  Here is a modified version of my original example with -regress-:
>
> sysuse auto, clear
> gen mpgturn=mpg*turn
>
> regress price mpg turn mpgturn
> forvalues x=31(1)51 {
>         nlcom _b[mpg] + _b[mpgturn]*(`x')
> }
>
> regress price mpg turn c.mpg#c.turn
> margins, dydx(mpg) at(turn=(31(1)51))
>
>> And the second thing is to ask if there is a way to easily plot those results
>> such as marginsplot does with margins saved output.
>
> Maybe someone else has some suggestions here...?
>
> --Mark
>
>
>> Any suggestions would be very much appreciated, as I am facing the same
>> need as Jana in the original question, i.e. I need to generate a marginal
>> effects plot for an interaction between the endogenous explanatory variable
>> and a exogenous explanatory variable (both continuous).
>>
>> Thanks,
>>
>> Suyin
>>
>>
>>
>>
>>
>> 2013/12/26 Schaffer, Mark E <M.E.Schaffer@hw.ac.uk>:
>> > Jana,
>> >
>> > -ivreg2- doesn't (yet - we really need to add this in!) support -margins-.  But
>> an easy workaround is to use -nlcom- with -ivreg2-.
>> >
>> > Below is an example of how to use -nlcom- to mimic -margins- to calculate
>> marginal effects and CIs when you have an interaction.
>> >
>> > HTH,
>> > Mark
>> >
>> > sysuse auto, clear
>> > gen mpgturn=mpg*turn
>> > regress price mpg turn mpgturn
>> > qui sum turn, meanonly
>> > nlcom _b[mpg] + _b[mpgturn]*(`r(mean)') qui sum mpg, meanonly nlcom
>> > _b[turn] + _b[mpgturn]*(`r(mean)') regress price mpg turn c.mpg#c.turn
>> > margins, atmeans dydx(*)
>> >
>> >> -----Original Message-----
>> >> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
>> >> statalist@hsphsun2.harvard.edu] On Behalf Of Jana von Stein
>> >> Sent: 26 December 2013 01:59
>> >> To: <statalist@hsphsun2.harvard.edu>
>> >> Subject: st: Re: Marginal effects after ivreg2
>> >>
>> >> Hello,
>> >>
>> >> I need to run margins (or calculate marginal effects with ci's) after
>> >> ivreg2. I have an endogenous interaction and want to calculate the
>> >> marginal effect (w/
>> >> ci's) after ivreg2. Does anyone know how to do this? Margins did not
>> >> seem to work after ivreg2.
>> >>
>> >> Thanks, and happy holidays.
>> >> Jana
>> >>
>> >>
>> >>
>> >>
>> >> *
>> >> *   For searches and help try:
>> >> *   http://www.stata.com/help.cgi?search
>> >> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> >> *   http://www.ats.ucla.edu/stat/stata/
>> >
>> >
>> > -----
>> > Sunday Times Scottish University of the Year 2011-2013 Top in the UK
>> > for student experience Fourth university in the UK and top in Scotland
>> > (National Student Survey 2012)
>> >
>> > We invite research leaders and ambitious early career researchers to
>> > join us in leading and driving research in key inter-disciplinary themes.
>> > Please see www.hw.ac.uk/researchleaders for further information and
>> > how to apply.
>> >
>> > Heriot-Watt University is a Scottish charity registered under charity
>> > number SC000278.
>> >
>> >
>> > *
>> > *   For searches and help try:
>> > *   http://www.stata.com/help.cgi?search
>> > *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> > *   http://www.ats.ucla.edu/stat/stata/
>
>
> -----
> Sunday Times Scottish University of the Year 2011-2013
> Top in the UK for student experience
> Fourth university in the UK and top in Scotland (National Student Survey 2012)
>
>
> We invite research leaders and ambitious early career researchers to
> join us in leading and driving research in key inter-disciplinary themes.
> Please see www.hw.ac.uk/researchleaders for further information and how
> to apply.
>
> Heriot-Watt University is a Scottish charity
> registered under charity number SC000278.
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/