3 ms·

Median is a tough number to use if a bunch of folks at the high end skew the data during a month- especially if volumes are drying up. To get at what is occurr

Median is a tough number to use if a bunch of folks at the high end skew the data during a month- especially if volumes are drying up. To get at what is occurring you probably need to show it in price bands. say top/bottom and middle third of real estate prices. How much volume at each and trends in each.

nickm12 10y ago
You have it reversed. The advantage of the median (50th percentile) is that a few outliers at the high end won't skew the data. To move the median, you have to move half of the distribution.
- jboydyhacker 10y ago
  No read the second sentence. What I'm saying is the real estate market has very different characteristics in the low, medium, and high end. So the top third, middle thirds and bottom thirds should be broken out and you can't use median for the market as one sample set.
rgbrgb 10y ago
We actually explored this a bit in our last post: https://www.openlistings.com/blog/sorry-sf-housing-prices-are-only-falling-for-the-one-percent/ https://www.openlistings.com/blog/sorry-sf-housing-prices-ar... I agree that it can be misleading but it's the most common statistic used to describe the health of a real estate market. I think that's because it's rather robust to outliers. But, as we show in the post above, it's not robust to tightening / loosening of the price bands above and below median (e.g. median went down last month but the 1st quartile went up).
- Dr_tldr 10y ago
  But isn't there a little bit of circular reasoning there? The median is the best statistic because it's the most common, and it's the most common statistic because it's the best. I'm not advocating for any other single-number analysis presentation, nor do I think the median number is totally meaningless, but you're leaving a lot of data out in the name of creating a simple chart without asking if it's informative.
  - rgbrgb 10y ago
    Yes, it would be circular had I concluded that median is best.

nickm12 10y ago

You have it reversed. The advantage of the median (50th percentile) is that a few outliers at the high end won't skew the data. To move the median, you have to move half of the distribution.

jboydyhacker 10y ago
No read the second sentence. What I'm saying is the real estate market has very different characteristics in the low, medium, and high end. So the top third, middle thirds and bottom thirds should be broken out and you can't use median for the market as one sample set.

rgbrgb 10y ago

We actually explored this a bit in our last post: https://www.openlistings.com/blog/sorry-sf-housing-prices-are-only-falling-for-the-one-percent/ https://www.openlistings.com/blog/sorry-sf-housing-prices-ar... I agree that it can be misleading but it's the most common statistic used to describe the health of a real estate market. I think that's because it's rather robust to outliers. But, as we show in the post above, it's not robust to tightening / loosening of the price bands above and below median (e.g. median went down last month but the 1st quartile went up).

Dr_tldr 10y ago
But isn't there a little bit of circular reasoning there? The median is the best statistic because it's the most common, and it's the most common statistic because it's the best. I'm not advocating for any other single-number analysis presentation, nor do I think the median number is totally meaningless, but you're leaving a lot of data out in the name of creating a simple chart without asking if it's informative.
- rgbrgb 10y ago
  Yes, it would be circular had I concluded that median is best.