This is weird…I went back and duplicated the regression Rich describes above on the RB dataset I mentioned previously. My sample size is much smaller, but maybe RB is different.
I got an r-squared of 75% and a coefficient that would imply for every increase of 100 square feet, the $/square foot goes down by $11, or for every 1,000 increase in square footage, $/square foot goes down by $110 (obviously, this would change in the limit).
My dataset is much smaller and only consists of homes <600K in the RB/4S area, but I'm still surprised to get such a high coefficient when the impact for SD in total is negligible.
Rich if you have a minute and it's easy, I'd be interested to see if you reach the same conclusion if you limit the dataset to detached houses under $600K.
