The worst statistical summary ever?

[Chris Lloyd]

A recent article in the Age*pointed out that Melbourne’s dams are approaching 50% capacity for the first time in 4 years.* They included a time series of % storage for the past 40 years which tells the whole story pretty well, placing the last four years of drought in context, as well as the recent rains. Unfortunately, they added a statistical summary which, in my opinion, ranks with the worst ever*.

They listed a time series of the number of days during that year where capacity was less than 50%. Not surprisingly, for many wet years there we no days under 50%. So there are lots of zeros. It’s almost as*dumb as summarising my growth curve by listing the number of days during each year*of my life I was over 5 feet tall.**Bearing in mind how highly correlated the underlying measurement is, why would you ever try to summarise it by looking at the number of consecutive values that are in a range?* If you were going to move away from the mean, the yearly high and low would be much more informative.

But it gets worse. They indicated on the table when new dams were opened, including the humungous Thompson dam.* Now when a dam opens it is surely empty. The state Premier cuts the ribbon when the wall and sluice gates are complete. And then the water flows in*over a number of years. So with a huge damn like the Thomspon which accounts for most of Melbourne’s capacity, you would drop below 50% on the day it opens and for a few years after that while it fills. But the table doesn’t show this. In fact, the opening of the Thompson dams does not seem to affect their daft statistic “number of days under 50%” much at all. So they maanged to add a confusing bit of extra information to a meaningless list of numbers.*

*Unfortunately, the graphic I am describing is not on the link. It only appeared in hard copy.



Get More from the original blog...