Hacker News Reader

Comment by aleyan | Hacker News Reader

This article spent paragraphs describing a normal distribution where a couple of descriptive statistics would do (mean and standard deviation). Ok, I kid, that would be too technical. The old adage still stands though: a picture is worth a thousand words. Instead of showing us what this curve looks like, they wasted the one graphic of their column inches on a useless stethoscope Gordian knot illustration.

One a plus side, this article appears to be famous and from 2004. Somebody must have produced graphic of this bell curve of CF outcomes by hospitals. Anybody got a link?

View on HN

...we know what the graph would look like. the author knows what the graph would look like, the editor knows what the graph would look like.

The article isn't about the curve proves, the article is about what the curve means.

(it also says in an early paragraph that you will see different distributions for different diseases/operations; there isn't just one curve)

I know what a normal distribution looks like, but I have seen too many flaws in data or flaws in characterization of data revealed by graphs of data. If this data and its graph is so earth shaking, show it.

Worse than that: if all hospitals were functioning perfectly, we would expect to see exactly a bell curve due to small random variations adding up.

A bell curve is not evidence that anything is wrong, and in fact it could be taken as evidence that everything is fine: a bell curve suggests that no hospital is doing remarkably badly.

So this part of the article makes it into the headline despite telling us nothing, and reduces the credibility of the whole thing.

I actually wonder if it really is a normal distribution.

A lot of people think any distribution with spread is a "bell curve". In a lot of fields like this the tails are a lot fatter than they should be in a normal distribution.

And doctors (author is a surgeon) don't get much statistical training. I was once told by a doctor that all random distributions are normal curves.

I also wanted more visuals or highlighted quotes. I opened the article, scrolled through it, saw there was no high-level answer to the titular question and closed it. I'm curious, but don't have time to read the article. ¯\_(ツ)_/¯

Reply on news.ycombinator.com