--- In evolution-discussions@yahoogroups.com, "J.Toby Mordkoff"
<jtm12@p...>
wrote:
> With that said, the values appear to be a column of numbers that
> subtract 0.127 from the previous values, which might give them a mean
> of zero, but does not alter the variance.
Close, we were just substracting the median. This results in half of the
classes appearing tougher and half appearing softer, which makes a certain
amount of sense. But the scale is fairly meaningless, and certainly cannot
be compared year to year.
Here's the proper statistical normalization, since J. Toby shamed me into
it. The order is the same; the classes at the top got tougher and the
classes at the bottom got easier. The scale is more meaningful, but it has
the odd notion that more classes got softer than tougher since it is based
off the mean rather than the median.
CLASS 2003 2004 %Diff Normalized Diff
STS 0.784 0.791 -0.885 -3.421
SM 0.834 0.838 -0.477 -2.032
ASP 0.846 0.848 -0.236 -1.209
DSP 0.818 0.819 -0.122 -0.822
BS 0.825 0.826 -0.121 -0.818
CP 0.853 0.854 -0.117 -0.805
FP 0.861 0.862 -0.116 -0.801
AS 0.828 0.828 0.000 -0.406
ES 0.804 0.804 0.000 -0.406
AM 1.000 1.000 0.000 -0.406
F125 0.938 0.938 0.000 -0.406
BM 0.946 0.945 0.106 -0.045
DP 0.845 0.844 0.118 -0.002
BSP 0.840 0.839 0.119 0.001
SS 0.837 0.836 0.120 0.002
DS 0.795 0.794 0.126 0.024
GS 0.782 0.781 0.128 0.031
BP 0.865 0.863 0.232 0.384
EP 0.863 0.861 0.232 0.386
SM2 0.851 0.849 0.236 0.397
CSP 0.839 0.837 0.239 0.409
ESP 0.826 0.824 0.243 0.421
FSP 0.815 0.813 0.246 0.433
HS 0.778 0.776 0.258 0.473
CM 0.916 0.913 0.329 0.714
AP 0.871 0.868 0.346 0.772
CS 0.814 0.811 0.370 0.855
FS 0.806 0.803 0.374 0.867
STX 0.798 0.795 0.377 0.880
EM 0.899 0.894 0.559 1.500
FM 0.894 0.889 0.562 1.511
DM 0.891 0.886 0.564 1.517
> I think the deeper point is that just because your PAX went up
> (down), it doesn't mean that your class is now much harder (easier)
> than most of the others ... just that A-mod got faster and couldn't
> be changed.
Right.
Mark
|