Statalist


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

st: marginal effects


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   stata list <statalist@hsphsun2.harvard.edu>
Subject   st: marginal effects
Date   Mon, 9 Mar 2009 19:55:50 +0000 (GMT)

--- Mark W. Smith wrote to me privately:  
> Puneet Singhal's example uses a log link.  Do you think it would be best
> to specify "mfx, dyex" rather than "mfx"  (= "mfx, dydx")?   Suppose
>  that a regular regression model (no log link) can be written as
>  
> y = x'B. 
>  
> Of course dy/dx = B.
>  
> In the GLM model with log link,
>  
> E[y|x] = exp(x'B)
>  
> and so dy/dx does not equal B. 
>  
> This is why I worry that using "mfx" (= "mfx, dydx") is not correct for
> a model with a log link.

The point of -mfx- is to get effects that are not equal to B, exactly 
because of the mechanism you specify. So, -mfx, dydx- will produce the 
first derivative/slope/effect of the curve E[y|x] at the specified 
levels of x (by default the means if you don't specify the -at()- 
option). Therefore, I think that -mfx, dydx- is the correct 
specification for this situation.

Hope this helps,
Maarten

-----------------------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


      

*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index