Hi Peer
Your -tabstat xb, by(edu)- command gives you the mean predicted
probabilities. The probabilities reported by -estat mfx-, however, are
calculated at the means of the independent variables. These will not
be the same in general since the probabilities are a nonlinear
function of the attributes.
Hope this helps.
Arne
2009/1/10 Peer Skov <[email protected]>:
> Hi listers,
>
> Upon running an asclogit I use estat mfx to obtain marginal effects.
> When I do so stata also reports the probability of the choice for each alternative.
> why do these differ from the average over alternatives when using predict xb?
>
> asclogit choice lifetime2 sigp sige, case(id) alt(edu) noconst
> estat mfx
>
> predict xb
> tabstat xb, by(edu)
>
> Best,
> Peer
>
>
>
>
>
>
>
>
> ----------------------------------------
>> Date: Sat, 10 Jan 2009 15:49:25 +0000
>> From: [email protected]
>> To: [email protected]
>> Subject: Re: st: A logit approach that allows alternative specific variables
>>
>> Hi Peer,
>>
>> This should do it:
>>
>> asclogit choice lifetime2 sigp sige, case(id) alt(edu) noconst
>> gen hold = lifetime2
>> replace lifetime2 = lifetime1
>> predict xb
>> replace lifetime2 = hold
>>
>> Hope this helps.
>>
>> Arne
>>
>> PS It's better to address the whole list when you have a new question,
>> even when it's related to a previous one, as the whole collective of
>> Statalist users is much more likely to know the answer than one single
>> user.
>>
>> 2009/1/10 Peer Skov :
>>> Hi Arne,
>>> I have encountered yet another challenge with the asclogit procedure.
>>> As explained previously I model education choice where the covariates
>>> are lifetime income and two types of income risk.
>>> I have two competing measures of lifetime income and in order to illustrate a point
>>> I would like to estimate the parameters using lifetime income measure 2
>>> but then obtain the predicted probabilities using lifetime income measure 1.
>>> That is, something along the lines of:
>>>
>>> asclogit choice lifetime2 sigp sige, case(id) alt(edu) noconst
>>> predict xb
>>> where x = [lifetime1]
>>>
>>> Do you have suggestions on how to do this?
>>> Best,
>>> Peer.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>> Date: Wed, 10 Dec 2008 12:43:23 +0000
>>>> From: [email protected]
>>>> To: [email protected]
>>>> Subject: Re: st: A logit approach that allows alternative specific variables
>>>>
>>>> Hi Peer,
>>>>
>>>> Some responses below:
>>>>
>>>>> I would like to specify the values at which the marginal effects are calculated.
>>>>> However when I use the at () option I specify the same life_earnings for all educational groups,
>>>>> is it possible to specify different levels for each group?
>>>>
>>>> I don't think so but I could be wrong. You'll probably have to
>>>> calculate the marginal effects manually if this is what you want,
>>>> Greene has the relevant formula for alternative-specific variables.
>>>>
>>>>> In addition, I was hoping to be able to produce a figure for each educational level depicting prob(edu=1) on the y-axis
>>>>> and income on the x-axis. That is, similar to Greene fig. 19.2 (Econometric Analysis, Fourth Edition).
>>>>> Do you know of any useful commands in that regard?
>>>>
>>>> I don't have my copy of Greene at hand but are you thinking about
>>>> something like:
>>>>
>>>> webuse choice
>>>> asclogit choice dealer, casevars(sex income) case(id) alternatives(car)
>>>> predict p if e(sample)
>>>> scatter p income if car == 1
>>>>
>>>> Arne
>>>> *
>>>> * 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/
>>> _________________________________________________________________
>>> Show them the way! Add maps and directions to your party invites.
>>> http://www.microsoft.com/windows/windowslive/events.aspx
>>> *
>>> * 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/
> _________________________________________________________________
> Invite your mail contacts to join your friends list with Windows Live Spaces. It's easy!
> http://spaces.live.com/spacesapi.aspx?wx_action=create&wx_url=/friends.aspx&mkt=en-us
> *
> * 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/
