nisse at lysator.liu.se
Thu Dec 10 17:22:44 CET 2009
Torbjorn Granlund <tg at gmplib.org> writes:
> Perhaps I should be this on geometric mean instead, or some combination
> of arithmetic and geometric mean?
I hope you first compute the ration measured/donfigured for each
machine, and then take the average of those values? For the usual case,
I wouldn't expect much difference between geometric and arithmetic mean.
In the case of SET_STR_DC_THRESHOLD, some of the values are clear
outliers (meas. 1085, cfg. 3009). If for some reeason you want to filter
these out (but they should indicate either some serious misconfiguration
or problems with tuneup), it would make more sense to me to switch to
thhe median than to a geometric mean.
> OK, so now brown means the configured value is to low ("down in the
> earth"), blue means the configured value is too high ("up in the sky")
> and green means it is right...
It might also be relevant to list maximum and minimum values for the
meas/cfg ratios, and use a separate color for threasholds with values
both far above and and far below 1 (these could end up as green if the
deviations happen to cancel when you compute the mean).
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
Internet email is subject to wholesale government surveillance.
More information about the gmp-devel