Le mercredi 15 février 2006 à 15:37 -0500, Morten Welinder a écrit : > > OK, I see, it is quite easy to implement. Just two more questions: > > - should I use the same x_i values when evaluating the smoothed ones (I > > guess the answer is yes). > > You can use any x value for m(x), including the x_i values and anything > in between. In fact there is no need to you splines for interpolation.
I don't use splines for smoothed curves, only for regressions. > > - how should I compute a default value for h ((last_x - first_x) / 10 ?) > > For the "line" case, use (say) 10 * the _step_ you chose between x values. > For the "xy" case, use something like 10 * (last_x - first_x) / N. The smoothing code does not know what type of plot it is working with (it can of course know the name of the plot type and if the x-axis is discrete or not). It just get two arrays filled with the x and y values. In the xy case, we also might have disordered x values, which makes things more difficult. For the moving average smoothing, I duplicated xl behavior it that case. I feel I'll use 10 * (last_x - first_x) / (N - 1) for all cases and skip invalid values where N is the number of original data points. Jean _______________________________________________ gnumeric-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/gnumeric-list
