bayes: mlogit - error

[Alberto Cardaci]

Hello everyone,
I have the following vector (A) with 73 observations:


A | Freq. Percent Cum.
-----------+-----------------------------------
ob | 2 2.74 2.74
lb | 57 78.08 80.82
pb | 3 4.11 84.93
unb | 11 15.07 100.00
------------+-----------------------------------
Total | 73 100.00


I want to estimate the predicted probability for each of the four outcomes (ob, lb, pb, unb) by running an intercept-only bayesian multinomial logistic regression. I type "bayes: mlogit A" and I get the following error:

new vector must be 1 x 3
r(198);

Any idea why I get this? NB: standard "mlogit A" command works fine.

Any help is really appreciated! Thanks a lot

[Carlo Lazzaro]

Alberto:
works for me in the following toy_example:

. set obs 10
number of observations (_N) was 0, now 10

. g A=1 in 1/4
(6 missing values generated)

. replace A=2 in 5/6
(2 real changes made)

. replace A=3 if missing(A)
(4 real changes made)

. label define A 1 "bad" 2 "medium" 3 "good"

. label val A A

. bayes: mlogit A
  
Burn-in ...
Simulation ...

Model summary
------------------------------------------------------------------------------
Likelihood: 
  medium good ~ mlogit({medium:_cons},{good:_cons})

Priors: 
  {medium:_cons} ~ normal(0,10000)
    {good:_cons} ~ normal(0,10000)
------------------------------------------------------------------------------

Bayesian multinomial logistic regression         MCMC iterations  =     12,500
Random-walk Metropolis-Hastings sampling         Burn-in          =      2,500
                                                 MCMC sample size =     10,000
Base outcome: bad                                Number of obs    =         10
                                                 Acceptance rate  =      .4293
                                                 Efficiency:  min =      .1445
                                                              avg =      .1504
Log marginal likelihood = -20.177187                          max =      .1562
 
------------------------------------------------------------------------------
             |                                                Equal-tailed
             |      Mean   Std. Dev.     MCSE     Median  [95% Cred. Interval]
-------------+----------------------------------------------------------------
medium       |
       _cons | -.8180582   .9518794   .025037  -.7803201  -2.909695   .8691686
-------------+----------------------------------------------------------------
good         |
       _cons |  .0066208   .7536667   .019067   .0157107    -1.4501   1.475032
------------------------------------------------------------------------------
Note: Default priors are used for model parameters.

.

[Alberto Cardaci]

Dear Carlo,
hanks a lot for your reply. Your example works for me. But my code still doesn't. Here's what I do:

label define Al 5 lb 9 unb 6 pb 2 ob
label values A Al
bayes: mlogit A

Again, I get:

new vector must be 1 x 3
r(198);

Here's the vector A that I'm using: https://www.dropbox.com/s/w31ebyunnd...ector.dta?dl=0

I'm using Stata 15.1.

Thank you!

Alberto

[Carlo Lazzaro]

Alberto:
what if (proposed amendments in red):

label define A 5 "lb" 9 "unb" 6 "pb" 2 "ob"
label values A Al
bayes: mlogit A

If you want to share your data with interested listers, please use -dataex-. Thanks.

[Alberto Cardaci]

Dear Carlo,
Sorry, I didn't know how to share files. Thank you very much for your suggestions!
Apparently the problem was related to the values of my vector A: changing the numbers from 5,9,6,2 to 1,2,3,4 solves the problem. I still don't get the reason behind this, but it all works now.
Thanks,
Alberto

[Carlo Lazzaro]

Alberto:
see -help dataex- (data sharing with -dataex- is easier than it would seem at a very first glance).