Hacker News Reader

Comment by averagewall | Hacker News Reader

Oh no, not another graph with misleading axes. It doesn't really show such a perfect fit as it looks like. They've both scaled and translated the axes to make the two curves look similar to each other. If they counted cars upwards from 0 instead of downwards from the previous amount, then it'd make more sense but then we would see if the slopes were different or there was an offset indicating, for example, that cars are expected to still remain after the stock price reaches 0.

Good on them for starting the stock price axis at 0, that's a rarity in stock-price graphs.

View on HN

Of course one of the axes has been scaled. We're interested in the proportionality between the two variables, to see how good their claim that cars parked is a good proxy measure for stock price. I see where "percent since arbitrary start time x" is a poor choice of units for cars, but since it's a linear transformation, you'd get the same picture (and conclusion) were you using "thousands of cars" instead.

I disagree that it shows proportionality. Since we don't know where the zero would be with "thousands of cars", it could be way off the bottom of the chart, which means car numbers are relatively constant, even as stock price falls to a few percent of it's starting value.

The graph does show that both variables are decreasing, and that they both have a knee in the middle where the rate slows. But that's about all they clearly have in common.

An yes, scaling is of course necessary. It's translating that's the problem. I shouldn't have said scaling in my first post.

Sorry, I still disagree with that comment. They're not comparing apples and oranges, here, they're saying "here's why we're using apples to approximate oranges". The graph is showing a close linear relationship between # of cars and stock price.

It's saying "There is some a and b such that:

    a * "number of cars" + b ~ "stock price"

You can have a gripe with that method, sure, but the graph is fine.

Reply on news.ycombinator.com