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]

From |
Arne Risa Hole <arnehole@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: MIXLOGIT: marginal effects |

Date |
Mon, 6 Feb 2012 17:04:41 +0000 |

Thanks Richard, these are all good points. To be slightly pedantic I did say 1 _unit_: height, for example, can be measured in metres or centimetres (or feet and inches) and the procedure I outlined will clearly be more sensible if the latter is used. Arne On 6 February 2012 15:08, Richard Williams <richardwilliams.ndu@gmail.com> wrote: > At 08:18 AM 2/6/2012, Arne Risa Hole wrote: >> >> Dear Davide, >> >> I haven't created a module for calculating marginal effects after >> -mixlogit- but this can be done using simulation as follows: >> >> 1) Use -mixlpred- to calculate predicted probabilities in the base >> scenario >> >> 2) Increase the relevant variable by 1 unit for all individuals in the >> sample >> >> 3) Use -mixlpred- to calculate predicted probabilities in the >> alternative scenario >> >> 4) Calculate the difference between the predicted probabilities in 1) >> and 3) and average this difference over individuals. This gives you an >> estimate of the marginal effect. > > > I've never used -mixlogit- so Arne can correct me if I am wrong, but I would > modify the advise as follows: > > The "increase by one" part may or may not work well; it depends on the > variable scaling. It might work well if the variable ranges between 0 and > 10,000 and not work well if it ranges between 0 and .10. It might be better > to increase by, say, .001, and divide the difference by .001, i.e. divide > the change in Y by the change in X. > > P. 353 of "Microeconomics Using Stata, Revised edition" recommends (p. 353, > section 10.6.10) using a change equal to the standard deviation of the > regressor divided by 1,000. The book can be bought from the Stata bookstore > and the code can be downloaded for free. > > Also, the procedure will differ for continuous and discrete variables. With > discrete, you would first set the value to 0, then set the value to 1, and > compute the difference. For an example see slides 28-30 of > http://www.nd.edu/~rwilliam/stats/Margins01.pdf . > > > ------------------------------------------- > Richard Williams, Notre Dame Dept of Sociology > OFFICE: (574)631-6668, (574)631-6463 > HOME: (574)289-5227 > EMAIL: Richard.A.Williams.5@ND.Edu > 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/ * * 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/

**References**:**Re: st: MIXLOGIT: marginal effects***From:*Arne Risa Hole <arnehole@gmail.com>

**Re: st: MIXLOGIT: marginal effects***From:*Richard Williams <richardwilliams.ndu@gmail.com>

- Prev by Date:
**Re: st: interaction effects vs. interaction coefficients using contrast?** - Next by Date:
**Re: st: Re: Nearstat version** - Previous by thread:
**Re: st: MIXLOGIT: marginal effects** - Next by thread:
**Re: st: MIXLOGIT: marginal effects** - Index(es):