" and take enough
counts to look for any pattern. "

When I look at that graphed data what I see is 15 points where the sticky board count is very low but the wash count is very high compared to the sticky board.  We have a fair idea on the error bar on the wash counts  and that error bar does not come close to telling us why these points are so far off the line.  Thus the only conclusion I can draw is the error bar on the drop counts is huge.  There are also points where the drop counts are way too high compared to the wash counts and again the error bars on the washes will not come close to explaining the deviation.  It looks to me like over 1/3 of the drop counts are so in error as to be meaningless and at least 24 hour drops are not worth doing before treatment.  I can see that 24 hour drop counts after treatment might tell you a lot about the effectiveness of a treatment.

Ordinarily I would suggest doing a least squares data fit rather than eyeballing the line.  In this case a least squares fit would give a line with less slope and is probably no more meaningful than the eyeball line.  I say that as the eyeball line tends to ignore the bad outliers and those data points seem to likely have some kind of systematic error and probably should be ignored.  Many of those outliers are bad enough due to either high or low drop counts that are so extreme compared to the paired wash counts that they can not be correct  Standard statistical tests would throw those data points out I expect, so ignoring them is proper. But when you throw out over 1/3 of your data set you probably better not trust any conclusion you might draw.

Dick

             ***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software.  For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html