On Fri, 22 Jul 2011, amaryl wrote: > I'm running a qlrtest for a structural break in a simpel OLS-regression > with a constant and two dependend variables. gretl gives me a > break-date, significant at the one-percent-level, the critical value of > which is given as 6.02. gretl gives as source of those critical values > Stock and Watson (2003): Introduction in Econometrics. > I take it, those asymptotic critical values for a qlr test were > originally calculated by Andrews (1993) and corrected also by Andrews > (2003). I just checked with the Andrews (2003) paper and if I read that > table correctly, the critical value of a qlr test with 15% trimming and > 3 degrees of freedom is 18.07. Yes, Andrews is the source for the values given in Stock and Watson. But the values are consistent. Andrews gives critical values for chi-square(q) whereas the S&W approach, used in gretl, is to compute an F-test. The critical values for the latter are the values given by Andrews divided by q. Allin Cottrell _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users