1. Firstly I run my mlogit model...
mlogit euivar /*
> */a16_24 a25_34 a50_spa notbrnuk nonwht hiq_deg hiq_hed hiq_alev hiq_gce hiq_oth /*
> */married marsep yeskid buying private cncloth autumn winter spring coh04 if male==1 & NWdum==1, rrr
2. Then I can find the sample proportions
>tab euivar if e(sample)
e, u, i | Freq. Percent Cum.
-------------+-----------------------------------
employment | 19,260 80.66 80.66
unemployment | 1,009 4.23 84.89
inactive | 3,608 15.11 100.00
-------------+-----------------------------------
Total | 23,877 100.00
3. I use the predict command as follows, and it suggest to me that the predicted probability is equivalent to the sample proportions (above) - e.g. a mean of 80.66 % is equivalent to the probability, 0.8066.
.predict p0 p1 p2 if e(sample)
(option pr assumed; predicted probabilities)
. su p0 p1 p2
Variable Obs Mean
p0 23877 .806634
p1 23877 .0422582
p2 23877 .1511078
4. However, using the prvalue command, the 'predicted probability' of employment is calculated to be higher (0.8764). I dont understand why.
.prvalue
mlogit: Predictions for euivar
Predicted probabilities for each category:
Pr(y=unemployx): 0.0295
Pr(y=inactivex): 0.0942
Pr(y=employmex): 0.8764
5. Similarly, it is possible to obtain the predicted probability using the following command, which gives the
'predicted probability' of employment=0.8764. Again, is this value different to both the sample proportion and the value given by the 'predict' command.
.mfx compute, predict(pr outcome(0)) nose
Marginal effects after mlogit
y = Pr(euivar==0) (predict, pr outcome(0))
= .87637938
________________________________
From: [email protected] on behalf of Richard Williams
Sent: Mon 05/09/2005 16:34
To: [email protected]
Subject: Re: st: Predicted probabilities after mlogit
At 10:10 AM 9/5/2005, Little, Allan wrote:
>Dear all,
>
>I'm trying to calculate predicted probabilities following a
>multinomial logit regression. My dependant variable has three states
>- employment (coded 0), unemployment (1) and inactivity (2).
>
>I would like to calculate the predicted probability, firstly
>evaluated at the mean, and secondly evaluated a number of different
>specified values. However, I am encountering some apparent
>inconsistencies according to which command I use.
>
>1. If I use 'predict p0 p1 p2 if e(sample)' [option 'pr' is assumed
>to be the default]. The predicted probabilities are equivalent to
>the proportion (sample mean) for each category.
Perhaps some output or at least the actual commands would help. Are
you saying that, for every case, the predicted probabilities are the
same for each category? I'm not totally clear on what you are doing,
since predict is going to give you values for individual cases
whereas prvalue gives you a few summary statistics.
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX: (574)288-4373
HOME: (574)289-5227
EMAIL: [email protected]
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
*
* 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/
<<winmail.dat>>
