Hacker News Reader

Comment by NikolaeVarius | Hacker News Reader

I've pretty much stopped paying attention to these numbers. I read/listen to multiple sources of news, and the amount of vastly different interpretations of these numbers has me concluding that its pointless to give a damn.

View on HN

Can you give an example of a single type of statistic where this argument can't be made? Are statistics overall categorically useless? Should they not be reported?

Statistics are certainly not categorically useless, but they are widely misunderstood and thus widely abused. In less political areas I have seen trained experts make elementary mistakes (mistakes happen and it's fine). A lot (nearly all in my informal sampling) of journalists have zero understanding of statistical methods. The term statistical significance leads to knee jerk reactions advocating policy without understanding the basic procedure, or implications of hypothesis testing.

As a numerate person I would personally appreciate it if all statistical claim had a link to a Jupyter Notebook and the associated data. (Though perhaps Observable would be the best choice here since it is run on the web).

I think in this case it’s worth pointing out that monthly unemployment numbers are historically very uncertain and the subsequent revisions to the initial number are often pretty substantial.

https://projects.fivethirtyeight.com/jobs-report-growth-unem...

This does not appear to be a very charitable interpretation of the above comment.

> Should they not be reported?

They probably shouldn't...

1) Humans are on average bad with numbers. This could have unforseen negative impacts.

2) Your average person can't define causal inference, let alone make a well-informed one.

3) Few of the presenters (news people) have a grasp on 1 & 2. If they do, also communicating that becomes a challenge that requires solving 1 & 2.

Reply on news.ycombinator.com