Producing Regression Table from bayesmh Using esttab

[afri afrimadona]

Dear Statalist members,

I would like to know how to produce a publishable regression table from bayesmh command. For example, I can easily produce a regression table from simple regress command like:


sysuse auto
foreach var of varlist mpg weight trunk length {
2. eststo: qui reg price `var'
3. }
(est1 stored)
(est2 stored)
(est3 stored)
(est4 stored)

. esttab, b(%9.3f) se(%7.3f) eqlabel(none) ar2 label

------------------------------------------------------------------------------------
(1) (2) (3) (4)
Price Price Price Price
------------------------------------------------------------------------------------
Mileage (mpg) -238.894***
(53.077)

Weight (lbs.) 2.044***
(0.377)

Trunk space (.. ft.) 216.748**
(77.146)

Length (in.) 57.202***
(14.080)

Constant 11253.061*** -6.707 3183.504** -4584.899
(1170.813) (1174.430) (1110.728) (2664.437)
------------------------------------------------------------------------------------
Observations 74 74 74 74
Adjusted R-squared 0.209 0.280 0.086 0.175
------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

.
end of do-file

.
But, I cannot do it after I run bayesmh command.


foreach v of varlist mpg weight trunk length {
2. eststo: qui bayesmh price `v', likelihood(normal({var})) prior({price: _cons `v'}, flat) prior({var}, jeffrey) saving(`v'model)
3. }
file mpgmodel.dta could not be opened
r(603);


Do you have any idea of how to produce a regression table from bayesmh command? Thank you



Best regards,


Afri

[Marcos Almeida]

Welcome to the Stata Forum/Statalist,

I believe that is not possible.

Besides, the SSC - esttab/estout - help files did not mention Bayesian analysis.

I cannot tell for sure about - bayesmh - (you may "save" results before using - estimates store, but not exactly the "regular" ones we see in the "frequentist" regression table). However, with - bayes - I gather the reason is the (quite different) items stored after the estimation.

By the way, even - estimates store - won't provide - estimates table - for that matter:

. bayes: reg price mpg
  
Burn-in ...
Simulation ...

Model summary
------------------------------------------------------------------------------
Likelihood:
  price ~ regress(xb_price,{sigma2})

Priors:
  {price:mpg _cons} ~ normal(0,10000)                                      (1)
           {sigma2} ~ igamma(.01,.01)
------------------------------------------------------------------------------
(1) Parameters are elements of the linear form xb_price.

Bayesian linear regression                       MCMC iterations  =     12,500
Random-walk Metropolis-Hastings sampling         Burn-in          =      2,500
                                                 MCMC sample size =     10,000
                                                 Number of obs    =         74
                                                 Acceptance rate  =      .4133
                                                 Efficiency:  min =      .1063
                                                              avg =      .1177
Log marginal likelihood = -727.25184                          max =      .1335
 
------------------------------------------------------------------------------
             |                                                Equal-tailed
             |      Mean   Std. Dev.     MCSE     Median  [95% Cred. Interval]
-------------+----------------------------------------------------------------
price        |
         mpg |  241.2524   21.16202    .62844   241.4937   198.4834   281.2732
       _cons |  49.63901   98.56165   3.02362   46.12188  -136.9005    248.258
-------------+----------------------------------------------------------------
      sigma2 |  1.58e+07    2723835   74553.2   1.55e+07   1.13e+07   2.20e+07
------------------------------------------------------------------------------
Note: Default priors are used for model parameters.
Note: Adaptation tolerance is not met in at least one of the blocks.

. estimates store mymodel

. estimates table mymodel
mymodel estimation results do not have e(b) and e(V)
r(321);

It seems the reason is stated in the last sentence.

[Marcos Almeida]

I just wish to add an extra comment: it seems the user-written - estout - was updated in 2014, hence, if I'm not wrong, the year before the release of Stata 14, when - bayesmh - was launched.

[afri afrimadona]

Thank you, Marcos. Yes, it seems that esttab does not anticipate this. In bayesmh, we use mcmc mean, mcmc SE, and so on rather than e(b) or e(V). I just wonder if there is a code that transfer them to esttab (or outreg maybe) or any way to transfer them into a standard regression table automatically.