>> I mean something like this... >> >> ================================================================ >> Spam Score -5 0 3 5 10 15 >> ---------------------------------------------------------------- >> Percentile 5.89% 12.59% 19.43% 24.81% 35.02% 50.37% >> ================================================================ >> >> just a thought. I'm normally thinking about what % of my mail hits a >> certain score (in that I'm worried about high scoring ham or low >> scoring spam. >> >> Regards, >> Rob > > Ah, got it... you like nice, integer spam scores, and interpolated > percentages. However, just so we're on the same page, your percentile > chart does not answer your question "what % of my mail hits a certain > score"... that's not the definition of percentile. The spam score 5 in > your chart maps to percentile 24.81, which means that 24.81 percent of > the values were *lower than score 5*, which of course is not the same as > saying score 5 was hit 24.81% of the time.
that's true, you need to consider the difference between the percentages to get an idea of that. I guess it could also be done as two tables. One for scores below 5 (or your spam/ham threshold) and another for those above. The important thing to me seems to be how much falls in the grey zones. ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ AMaViS-user mailing list AMaViS-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/