Hacker News Reader

Yeah I had to read this over 3 or 4 times before I understood exactly what the author was saying. Very unintuitive way of putting it.

View on HN

I think that a lot of statistical statements are just unintuitive and are very difficult to explain in terms people grok quickly.

Can either of you kindly help us understand in layman’s terms? (Or point to an informative URL resource?)

The way it makes sense to me is to invert the number you are looking at.

95 & 99 are very 'close together'

But if you look at 5 & 1 - the first number is 5 times the second. Huge relative difference

Sure, but the deal is that at very small numbers, a massive change in % difference doesn’t translate in to a massive difference. It’s _entirely_ dependent on context. The raw stats are meaningless.

The author is muddying the water for effect, that's why his explanation is confusing. To understand what he means, you should ignore the results expressed as percentages and instead focus on 1:19 vs 1:99.

Now, imagine a grid, where the unit on the x-axis is "No", vs "Yes" on the y-axis.

Place both studies in terms of their respective No/Yes relationship.

Study A: 1:19, 2:38, 3:47, ...

Study B: 1:99, 2:198, 3:297, ...

Can you picture the difference in their slopes?

Reply on news.ycombinator.com