Save specific output from multiple regressions

[John Andrews]

Dear all,

I am new to Stata and apologies if I post something in a wrong way.

I want to run many ARMA(p,q) models for p,q=0,1,3 and I want to gather together in one table the AIC and BIC values for all the combinations of the (p,q) values.

What I know so far is that to get the AIC and BIC for the ARMA(0,0) I type in
arima var, arima(0, 0, 0)
estat ic

But since I have many different ARMA(p,q) models is there a way to get the desired output on AIC/BIC in one table without seaching eah output separately?


I assume that the variables are stationary, so d=0 in ARIMA(p,d,q).

Best,
John
Stata 14SE
Windows 10

[Paul Dickman]

Welcome to the list! Here's some working code that I hope will get you started. I haven't fitted time series models since I was at university 30 years ago, so I took the example from the help file.

There are a lot of concepts for a new Stata user. For most of them it should be obvious where to look in the help (e.g., postfile, foreach).

A concept that is not so obvious is how Stata stores the results of calculations. Look in the help for -return-. The results from -estat- are stored in a 1 x 6 matrix where the 5th element is the AIC and the 6th the BIC. I write these to local macro variables.

If you type "ereturn list" after fitting the ARMA model you will see a list of stored results. If you want to save additional quantities then they can easily be added to the postfile.

version 14
tempname memhold
tempfile results

webuse wpi1
postfile `memhold' ar ma aic bic using `results', replace
forvalues ar = 0/2 {
forvalues ma = 0/2 {
      quietly arima wpi, arima(`ar',0,`ma')
      quietly estat ic
      local aic = r(S)[1,5]
      local bic = r(S)[1,6]
      post `memhold' (`ar') (`ma') (`aic') (`bic')
} 
}
postclose `memhold'
use `results'
list

Here is the output.

. list

     +-------------------------------+
     | ar   ma        aic        bic |
     |-------------------------------|
  1. |  0    0   1200.395   1206.036 |
  2. |  0    1   1039.044   1047.505 |
  3. |  0    2   885.7755   894.2363 |
  4. |  1    0   409.8548   418.3156 |
  5. |  1    1   353.2209   364.5021 |
     |-------------------------------|
  6. |  1    2   335.7601   349.8615 |
  7. |  2    0   305.1458   316.4269 |
  8. |  2    1   294.7223   308.8237 |
  9. |  2    2   295.7058   312.6275 |
     +-------------------------------+

[Unuabonah Eron]

please i am a new bee in stata and i really love how it works but right now i am faced with some challenges. i am working with sex as a categorical variable and i have successfully converted it using the code 'encode sex' but now i want to obtain the observation, mean and standard deviation of only the male sample. please what do i need to do? this is really pressing

[Unuabonah Eron]

furthermore i need help on how to obtain this file from my stata 14 software. sysuse auto.dta
(what i get in trying to obtain it is
'no; data in memory would be lost
r(4);'

[Paul Dickman]

There's a simpler solution to your specific question, than that I first posted.

version 14
webuse wpi1
forvalues ar = 0/2 {
forvalues ma = 0/2 {
      quietly arima wpi, arima(`ar',0,`ma')
      estimates store arma`ar'`ma'
} 
}
estimates stats arma??

Results are as follows:

. estimates stats arma??

Akaike's information criterion and Bayesian information criterion

-----------------------------------------------------------------------------
       Model |          N   ll(null)  ll(model)      df        AIC        BIC
-------------+---------------------------------------------------------------
      arma00 |        124          .  -598.1975       2   1200.395   1206.036
      arma01 |        124          .  -516.5222       3   1039.044   1047.505
      arma02 |        124          .  -439.8878       3   885.7755   894.2364
      arma10 |        124          .  -201.9274       3   409.8548   418.3156
      arma11 |        124          .  -172.6105       4   353.2209   364.5021
      arma12 |        124          .  -162.8801       5   335.7601   349.8615
      arma20 |        124          .  -148.5729       4   305.1458   316.4269
      arma21 |        124          .  -142.3611       5   294.7223   308.8237
      arma22 |        124          .  -141.8529       6   295.7058   312.6275
-----------------------------------------------------------------------------
Note: BIC uses N = number of observations. See [R] BIC note.

Using postfile is more general. The postfile code can be simplified.

version 14
tempname memhold
tempfile results

webuse wpi1
postfile `memhold' ar ma aic bic using `results', replace
forvalues ar = 0/2 {
forvalues ma = 0/2 {
      quietly arima wpi, arima(`ar',0,`ma')
      quietly estat ic
      post `memhold' (`ar') (`ma') (r(S)[1,5]) (r(S)[1,6])
} 
}
postclose `memhold'
use `results'
list

[John Andrews]

Thank you so much all of you. Perfect!!