From Newsgroup: comp.sys.mac.advocacy
Yes, that raw temp/kWh/day scatter plot showed a lot of randomness. But
as I tried to explain there were other variables that accounted for
apparent random scatter not explained by temperature. I just got around
to creating a scatter plot to show how that looks:
https://drive.google.com/file/d/1x5bPywVfOGnUdMoPhgFGM7rVXKNhBx9-/view?usp=sharing
As you layer in the home change variables in over time the pattern
emerges as a set lower and lower predicted kWh/day layers in the
predicted values. Thus the 97% R^2 for the final equation.
Notes: The windows were installed across 2 years. First the south side
and then north side of the house. The intercept shifter was 0.5 for the
first set and 1.0 for the second. Worked well.
There is no layer for days away from home. That remains as a random
shifter of predicted use.
--- Synchronet 3.21b-Linux NewsLink 1.2