Author: Chris Carson
Date: 20:01:56 01/27/00
Go up one level in this thread
On January 27, 2000 at 22:49:11, Dann Corbit wrote: >On January 27, 2000 at 22:43:22, Chris Carson wrote: >>On January 27, 2000 at 22:17:25, Dann Corbit wrote: >> >>>Have you done a correlational coefficient? >>> >>>It might also be interesting to calculate prediction and confidence intervals. >> >>Yes: >> >>Regression >>Analysis >>R = 1.00 Correlation Coef. >>SEM = 18.1 Error of measure (2 SEM = 95% confidence) >>Slope = 231.4 (number of SSDF points per 6x speedup, or per ply) >> (this means 77 points for 2x speed increase) >> (this would be for both hw and programming improvements) >>Y intercept = 272.8 >> >>Note: The regression analysis changes slightly as the SSDF continues >>testing (new lists), but not a whole lot (maybe 10% over time). > >I doubt the figures. A 1.00 correlation coefficient is a perfect correlation. >Something *must* be amiss. No measured data ever correlates perfectly. Regression Analysis R = 0.9971 SEM = 18.0602 Slope = 231.4427 Y intercept = 272.8312 At two significant digits the factor was 1.00, I expanded to two more digets out and as you can see, it is not a perfect one, however, please check my analysis, I am not perfect. :) A pure r=1.0 would yield a SEM=0, that was not the case. :) I can expand more digits, but I am not sure the additional accruacy is meaningful. :) Best Regards, Chris Carson
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.