Hello all, As you know, gretl automatically gives sequential numbers to models estimated in a given session. I think this is at least marginally useful, but it seems to me the details could be improved. The following script illustrates a couple of points that I think are problematic: <hansl> function matrix ols_coeff (series y, list x) ols y x return $coeff end function # first session open data4-10 ols 1 0 2 # "Model 1" series u = $uhat ols u 0 2 3 4 --quiet printf "LM: chi^2(2) = %.1f [%.4f]\n", $trsq, pvalue(X, 2, $trsq) ols 1 0 2 3 4 # "Model 3" # second session open data9-7 ols 1 0 2 # "Model 1" list X = 2 3 4 ols_coeff(QNC, X) # "Model 2" ols 1 0 2 3 4 # "Model 2" </hansl> In the "first session" the two models that are printed are numbered 1 and 3. It's true a model was estimated between those two, but I'm inclined to say that when a model is estimated with the --quiet flag numbering should be skipped. Then the two _visible_ models would be numbered 1 and 2. In the "second session" we see that within a function the numbering of models continues from the last number in the main script, but then gets reset on exit from the function. So we see two "Model 2" printouts. In this case I'd suggest that numbering of models be suspended in function calls. If a function writer wants to print a label for a given model, the "printf" command is available. In short, my proposal is that automatic numbering be applied only to "visible" models in the "main" portion of a gretl script. In many cases this will be the same as current behavior, but in more complex cases I think it would be an improvement. Any thoughts or objections? Allin Cottrell