You are viewing a single comment's thread.

view the rest of the comments →

[–] dadudemon1 0 points 4 points (+4|-0) ago 

Really appreciate this explanation. This line of yours is worth more than gold:

Positive bar for over-representation and negative bar for under-representation.

And that causes the whole thing to make sense.

I do stats as part of my job. Have done it for years. I can usually make heads or tails of shitty graphs in seconds. But this one had me scratching my head. But after your explanation, it makes so much damn sense that I feel embarrassed.

[–] NukeZionAlready 0 points 1 point (+1|-0) ago 

That's alright mate, it's the graph that sucks, you're not crazy ; a consistent mapping would be a 17-tall bar and then a 0.02-tall bar (1/50), or conversely a -0.06 bar and a -50 bar. But you can't just flip the sign around like they did. Otherwise there is a massive discontinuity around 1x representation: a 0.99 tall bar or -1.01 tall bar would be the same underlying value. You could theoretically do a log scale, with a +1.23 bar (log(17)) and a -1.7 bar (log(1/50)), but then normies wouldn't get it. They could also simply have labeled the axis "Representation ratio" for clarity and no one would be asking this question in the first place.

Someone should really do a better version of this graph...