Thanks, Scott! I'll do a net search on that. Unfortunately, it's taken me
a week to run the eight models from my survey data that I've now written
up. I think I'll have to drop all my reported MEs (which I
hand-calculated) from my tables, as I have little confidence that they are
correct. I'll run it on the quickest model to compute and see what
happens, and test out Michael Blasnik's prototype -estsave- command whilst
I'm at it.
By the way, unlike most Statapeeps on this list, I don't consider myself
loaded enough to afford all of the Stata manuals. I find it hard to
understand why people make this assumption. There are no books *at all* on
Stata available at my (rather large) university library, not least
because, quite wrongly in my view, Stata isn't on our university machines
here. I'll let you all into a dirty secret I have:
**I'm a student!**
I know. Incredible, isn't it?! Yes, I'd *love* to own the whole kit and
caboodle, but I had to make choices as to which I could purchase, since I
needed to purchase Stata 8 and Stat/Transfer 7. I chose to buy [U]. I'm
thinking of getting [XT] soon. But [R]? No way, not at that price. Love
to, but you know, finances and all that... ;-)
C.
> Another possibility, rather than calculating the marginal effects at the
> mean,
> you might find the average marginal effects more useful. Take a look at
> Tamas
> Bartus's -margin- command. This will be faster than -mfx compute- since
> it
> relies on analytical rather than numerical calculations.
>
> Hope this helps,
> Scott
>
>
>
> ----- Original Message -----
> From: "Clive Nicholas" <[email protected]>
> To: <[email protected]>
> Sent: Friday, November 14, 2003 5:51 AM
> Subject: st: On marginal effects
>
>
>> Nick,
>>
>> OK, no problem. I should reply by saying that I've attempted to
>> circumvent
>> this difficulty by tabulating Excel spreadsheets. However, consulting
>> the
>> literature has not made me entirely clear about the computation. As I
>> understand it, -dF/dx- as reported in Stata multiplies the probit
>> coefficients by their corresponding f(z) score. But when I downloaded
>> the
>> auto-data from Stata, my -probit- calculations of [b*f(z)] didn't match
>> the marginal effects presented after running both -mfx compute- or
>> -dprobit- (all of which were MEs of 14 zeros, anyway). Where am I going
>> wrong??!
>>
>> C.
>>
>
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
Yours,
CLIVE NICHOLAS,
Politics Building,
School of Geography, Politics and Sociology,
University of Newcastle-upon-Tyne,
Newcastle-upon-Tyne,
NE1 7RU,
United Kingdom.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
