Hi,

Sounds like you are losing your self in the processes and
confusing your goals. I usually need a visual representation
to sort out what I'm getting and what I actually want.

In this situation I would plot all of my data on an XY Scatter
chart, adjust my scales to something that looks usable
and then start analysing the data with dummy series and
trendlines. I'd plot dummy series showing the AVERAGE,
the GEOMEAN (even HARMEAN and TRIMMEAN sometimes
although I doubt they would be any good here). I plot them
as yearly series and monthly series of each maybe even daily
in some cases, also play around with some trendlines some
where in all this trial and error process I usually spot
a relationship in the data that is relevant to what I am
trying to achieve.

HTH
Martin


wrote in message
oups.com...
On Sep 22, 4:41 pm, "
wrote:
I want input on the validity (or not) of the year-over-year statistics
-- that is, the average and standard deviation of many sub-annual
year-over-year data in order to estimate annual statistics.

Well, I am obviously wrong-minded here, at least with respect to the
std dev. Imagine an example where there is significant volatility,
even on annual anniversary dates, but the day-to-day changes almost
identically track the day-to-day changes a year earlier. I might not
be explaining that very well, but the situation arose in a 4-year
subset of the actual data that I am working with. The std dev of the
year-over-year data is deceptively small.