On Mar 12, 7:28 am, "Iva"  wrote:

>I'm still using the older list just to keep those 3's down
>where they belong. 

Speaking as a researcher, I'm kinda amusing myself thinking up
different statistical techniques that could be used to determine a
contestant's "real" score. Any researcher worth their salt knows to be
suspicious when a range of scores is not normally distributed, and in
this case I rather suspect we'd see something like a trimodal
distribution with one cluster at zero, another cluster at 10, and
another cluster at the contestant's "real" score.

One method of analysis would be to calculate the mean, median and mode
of scores within the "middle" range. Another would be to look at the
degree to which each numerical score was contributory, across all
contestants -- for example, if '10' scores were found to contribute in
very few cases, they'd be considered to have a low correlation with
actuality and could be counted less in determining the contestant's
overall score. Yet another method would be to use an artificial normal
distribution to simulate how many of a particular score (e.g. 0 and
10) there ought to be, if such a score was determined to not be
realistic. (What I've listed here is by no means exhaustive, btw.)

So, if these corporate researchers had the sense to use these kind of
statistical methods, then votes of 8 or 9 could conceivably help Amy
more than votes of 10. Oh well, perhaps they're only market
researchers and not real scientists, so they might not think of these
things.

But in any case, Amy's been getting 10s from me regardless. They're
well deserved. :)

Matthew
--- SBBSecho 2.12-Win32