Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: margins after oprobit with interactions


From   Richard Williams <[email protected]>
To   [email protected]
Subject   Re: st: margins after oprobit with interactions
Date   Tue, 22 Nov 2011 08:31:58 -0500

At 10:16 AM 11/21/2011, Vincenzo Carrieri wrote:
Dear Statalisters,
I am estimating an ordered probit model:

Happy=age+age^2+income+income*age+income*age2 , that in "factor model style" is:

oprobit happy c.age c.age#c.age c.income  c.income#c.age c.income#c.age#c.age

where Happy is a 5 point scale oredered variable. I would like to have
marginal effects of all these variables. I am using the command
margins. So for instance, for the top category I write:

 margins , dydx(*) predict (outcome(5))

Stata gives me only two marginal effects: one for age and  one for
income. If I have understood well, these marginal effects are
estimated by STATA taking into account the fact that income is
interacted with age, age is squared and income is interacted also with
age squared. Is it right?

In addition, why I don't get also the marginal effect of the
interaction terms? If I do the first derivatives of the model, I get
dy/dincome, dy/dage but also dy/dagedincome. IS there a way to get
marginal effect for interaction terms with margins, or any other
command in Stata?

You can get marginal effects for interactions with the old -mfx- command. Unfortunately, they are wrong. You might want to read

http://www.stata-journal.com/article.html?article=st0063

Variations of this Q pop up regularly. To me, it makes no sense to want the marginal effect of an interaction term. The value of the interaction can't change without one of the terms used to compute the interaction changing. Any effect the interaction has is captured in the marginal effects for the terms used to compute the interactions. So, either there is no marginal effect for the interaction; or, if you do something like the above article does, you get a marginal effect for each case. But again, no single term. If I am wrong about this I'd love to have somebody point out the error in my thinking.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  [email protected]
WWW:    http://www.nd.edu/~rwilliam

*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index