On Sun, 26 Sep 1999, John Carmichael wrote: > Is the average EOT correction over the course of the entire year equal to > about seven or eight minutes? from (16+15)/2/2=7.75 min.. As 16 minutes is > about the extreme correction on Nov. 1st. and 15 minutes is the other > extreme corroction in mid Febuary. Does anybody know the exact average? > > The reason I ask, is that I state in my Sundial Owner's Manual that if one > does not have access to an EOT graph (or doesn't have it almost memorized, > as I'm sure many of you do), then, on the average, if you read the time > directly from a longitudinally corrected dial, your non-EOT corrected > reading would be off by about seven minutes on any day of the year. (Which > isn't bad as many clocks that I see are off by this much or more. > > OR, am I completely wrong; is the average EOT correction equal to zero?
Great question, John! This gets into the mathematics of what constitutes an "average". Your final comment is right in one sense - I suspect that if you add up (integrate) the EOT "steps" - say height*width - it will sum to zero. What the *user* percieves is the the dial (or clock, if you will) is "wrong" just about as much in February as it is in November. Sign doesn't matter! That's why when we do a "Least-Squares Fit" to match an equation to a set of data plotted on a curve, we add up the *square* of the error at each point, then take the average by dividing that sum by the number of points. That makes the sign disappear from the sum. From that point of view, you're right - the "typical error" is more like 7 or 8 minutes on any randomly chosen day of the year. Dave
