SB-
You’re correct to point out that median associated with a small sample size is sometimes unreliable. However, your subsequent assertion that a statistically accurate median requires all prices from all SFRs in the county is absurd.
Assuming an underlying normal distribution of home prices within any region, which is quite safe to do, standard statistical theories require only about 20 independent samples to reliably estimate the central tendency (i.e. the mean, median, or mode).
So yes, Chrispy and VRundy should be careful about reading too much of anything into the 30% drop in Delmar SFRs based on only 11 sales. At the same time there is useful information in some of the DQ numbers.
For instance, of the 43 sub-regions with greater than 20 SFR sales, 33 showed a YOY decline. To test the hypothesis that the market is flat, we assume that the chance of a decline or increase in any one region is equal to 50%. Given that assumption, the probability that one will see 33 out of 43 declining (or for that matter increasing) regions is 0.00021798 or roughly 2 out of 10,0000. So by all conventional criteria of statistical reliablity we can reject the hypothesis that the county-wide market is flat.
In addition, if we exclude all those regions with less than 20 sales and recompute the yoy change weighted by the number of sales in each of these ‘relaible’ regions, we see the mean drop for the county is closer to 3.37% YOY. Using the same back calculation including all the sales gives me a number close to the reported 2.2% YOY drop.
-tim
