Stored results of jb command

[Trung BUI]

Dear users,

I run the arque-Bera test for normality in stata. The results show as belows:

jb lnrgdp               //  Jarque-Bera asymptotic test for normality on the specified variable in level form.
Jarque-Bera normality test:  1.698 Chi(2)  .4279
Jarque-Bera test for Ho: normality:

Then, I used the command "return list" to show the results stored in the command. It shows like this:

 return list
scalars:
                  r(N) =  27
              r(sum_w) =  27
               r(mean) =  24.48843404981825
                r(Var) =  .2828294400632074
                 r(sd) =  .5318171114802601
           r(skewness) =  -.126910917299197
           r(kurtosis) =  1.798030007739949
                r(sum) =  661.1877193450928
                r(min) =  23.59275627136231
                r(max) =  25.30617713928223
                 r(p1) =  23.59275627136231
                 r(p5) =  23.66381454467773
                r(p10) =  23.71356582641602
                r(p25) =  24.0167236328125
                r(p50) =  24.50394058227539
                r(p75) =  24.96875953674316
                r(p90) =  25.19526290893555
                r(p95) =  25.24806213378906
                r(p99) =  25.30617713928223

I cannot find the scalars for Chi-square and p-value.

Are there anyone know how to get the value of chi-square (1.698 ) and pvalue (.4279) as what shown after running the command "jb"? Please help.

Thanks a lot in advance.

[Nick Cox]

There are two big problems here.

1. This is a terrible, irresponsible, absurd test despite its popularity among some economists. It uses asymptotic results for sampling distributions on small samples (size 27, in your case) for a problem in which convergence to asymptotia is appallingly slow. If you must use a test for normality, consider Doornik-Hansen or Shapiro-Wilk. Even better use qnorm

2. jb (from SSC, as you are asked to explain) is an old program written to less than current Stata standards which itself leaves no results in memory. What you see are just the results from its calling summarize. So what you want requires that you or someone else rewrite jb. But somewhere in official Stata there is a Jarque-Bera test. But see #1 again.

[Trung BUI]

Dear Cox,

Thanks for your advice.