Sven Schreiber

OK, here are the places that I found where it could make sense to mention Doornik-Hansen: -1- in lib/src/printout.c; in the function void print_freq_test (const FreqDist *freq, PRN *prn) this line: pprintf(prn, "\n%s:\n", _("Test for null hypothesis of normal distribution")); -2- in lib/src/graphing.c ; in the function int plot_freq (FreqDist *freq, DistCode dist, gretlopt opt) this line: fprintf(fp, "set label \"%s:\" at graph .03, graph .97 front\n", _("Test statistic for normality")); -3- in lib/src/gretl_model.c:...

adjust two German translations about normality tests

Well, for multivariate residuals (as with SYS and VAR) there is less ambiguity, I'd say - there cannot be a Jarque-Bera joint test for N>1, right? Also, for VARs the Doornik-Hansen test is mentioned in the user guide. In general it wouldn't hurt to mention the Doornik-Hansen fact also for single-equ models and for general systems in the user guide.

Sounds like a reasonable request to me. The fact that it's Doornik-Hansen is mentioned in the "modtest" reference, but not in the GUI, AFAICS. (Nor directly in the user guide, at least not for plain OLS models.) However, there's more than one string affected, namely also the one in the extra window when the test is chosen explicitly from the menus, and then perhaps also in the residual histogram plot.

So, can this be closed then?

I'm seeing various sub-aspects here: Information about the existence of oprobit_predict: Even for core gretl people like Jack it's not at all obvious that that feature already exists there. Another case for add-on status, such that hints can be added into the core gretl documentation? Or like Allin said, enhance the built-in stuff. In the multinomial case, the built-in probability results can only be saved in text format, not as a matrix. (In the GUI.) That's one example where oprobit_predict is...

add 2026 dev meeting folder

Selection of variables for list definition