On Fri, Feb 24, 2012 at 4:53 PM, Marc L. Allen <mlal...@outsitenetworks.com>wrote:
> You're trying to calculate it for individual people? Can you count on > night-time people to stay night-time, or do you need to worry about someone > shifting by 12 hours? > It's for individuals, and it is possible for individuals to shift or drift by any amount. > > If not, your best bet is, for the night-time people, add, say 6 hours to > all of their times, do your average, then subtract the 6 hours back out. > Yes, this is a good idea, the same as was given in another response. Thanks. I found that this type of measure is referred to as the "mean of circular quantities", and there is even a Wikipedia page about that...I had just never thought about it before. I also found the Mitsuta Method for dealing with this type of issue. But in any approach, things break down if data is strewn all over a 24 hour period. > There are cases where this will fail, but you might be able to detect data > sets that will cause this issue and ignore them. > I will have to just come up with a reasonable check of the data's variance and if I find it is all over the clockface, let the user know that the mean bedtime can't really be computed due to the erratic data. Maybe if only a few outliers are found I could filter them out. I may post a follow-up question regarding that. Thanks! Che _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users