On 8/19/2009 Pete Smith N4ZR wrote:
> This is exactly what was done in the VE5ZX/CT1BOH analysis. The trouble
> with it is that even with the really striking outliers, it's not direct
> evidence of any particular behavior that is contrary to the rules.
> That's why you need the possibility of inspection.
Agreed on all points here, even if the analysis was actually a fairly simple
analysis. In addition, there are techniques where you take a subset of the
data (say 20-25% - if you have enough data to be significant), you develop your
model on that, and then you run the model against the full data set
It's interesting what you can do with a small sample of even say 100 logs. It
would make a real interesting grad school project for someone taking statistics
(Hint to anyone from CQ Magazine reading - Stony Brook has a good stat program,
with students looking for Masters projects all the time)
73 de KG2V
For the Children - RKBA!
Clones are people two.
CQ-Contest mailing list