mlogit - estimate probability for specific value

[Truong Quoc Phan]

Hello stata community

I run the following command

mlogit y x1 x2 x3 i.year

Where y takes 3 values 0, 1, 2

Now I want to estimate the probability for a specific value of X. For example:
- x1=2, x2=5, x3=7, year=2000
- Estimate probability for y = 1

Then save the predicted probability to a scalar variable. I read the manual, there is a predict but don't know how to add the specific value for probability estimation

Thank you

[Clyde Schechter]

margins, predict(outcome(1)) at(x1 = 2 x2 = 5 x3 = 7 year = 2000)

To access the result and save it as a scalar

scalar wanted = r(b)[1,1]

[Jeff Pitblado (StataCorp)]

margins can do this for you.

To get the probabilities for each outcome, but without standard errors, try

margins, nose at(x1=2 x2=5 x3=7 year=2000)

The resulting probabilities are posted to matrix r(b).

To get the probability for y=1, try

margins, nose at(x1=2 x2=5 x3=7 year=2000) predict(outcome(1))

Here is a simple example using the auto data

sysuse auto
mlogit rep78 turn trunk displ
margins, nose at(turn=40 trunk=15 displacement=195) predict(outcome(1))
matrix list r(b)

Here is the output from margins.

. margins, nose at(turn=40 trunk=15 displacement=195) predict(outcome(1))

Adjusted predictions                                        Number of obs = 69

Expression: Pr(rep78==1), predict(outcome(1))
At: turn         =  40
    trunk        =  15
    displacement = 195

------------------------------------------------------------------------------
             |     Margin
-------------+----------------------------------------------------------------
       _cons |   .0017822
------------------------------------------------------------------------------

. matrix list r(b)

symmetric r(b)[1,1]
        _cons
r1  .00178223

[Truong Quoc Phan]

Jeff Pitblado (StataCorp)
Clyde Schechter

Thank you so much for your help
The solution works excellent

[Richard Williams]

The mtable command from Long and Freese’s spost13 routines is really great for producing nicely formatted output. See the first few pages of

https://www3.nd.edu/~rwilliam/xsoc73994/Margins05.pdf