Two random (for the purposes of this post) articles that show a clear lack of understanding of numbers and reason; or, careless oversight and a frivolousness use of words.

First, from this article on natural gas production:

In other words, the EPA thinks the amount of methane leakage is only about 80 percent of what was previously thought, despite the boom in natural gas production.

That clause at the end is either: a) unnecessary (if the author is using language sloppily), or b) shows a lack of understanding of what a percentage is because it doesn’t matter how much more methane we’re producing if the amount of leakage is expressed as a percentage of production.

And there’s this much more egregious error from an article on money “buying” happiness.

bad graph

 

Which the writers describe this way:

The next graph compares different countries, rather than different households within countries. Here, each circle represents a nation, with the richest ones clustered on the right. If extra income didn’t matter for well-being, you’d expect the line to flatten. Instead, it steepens.

What has happened here is that they’ve arbitrarily broken a single data set into two data sets. Why? I don’t know, but probably so that they could make a provocative conclusion: that income does matter for well being. Well, it’s bogus. At least, the graph is (the conclusion might not be, despite the bad data analysis, but that’s not my point). You can’t – reasonably – just arbitrarily break data sets into small data sets and form conclusions based on those. This is a common statistical error and leads to such “true” statements as 9 out of 10 dentists recommend Crest toothpaste or 9 out of 10 game-goers prefer Bud Light over Miller Light. What is one simple way to see that the dissecting of the above data set is bogus? Well, you could easily break the data set again at about 32 GDP per capita (the x-value) and then fit a line that goes very steeply downward. Your conclusion would then be that being very rich and getting richer is really bad for happiness. While this may or may not also be true, you can certainly understand why the authors did not also make this particular subdivision; it would have sharply contrasted with their desired conclusion.

-JD Cross

 

Advertisements