Hacker News Reader

Comment by drc500free | Hacker News Reader

Really nice work, but also quite misleading. The "real" map isn't interesting enough, so 97% of votes are excluded from the most populous liberal counties using an artificially low ceiling.

View on HN

As the author of the OP explains, those counties are shown with 100% saturation. Moreover, the author also shows a map without the upper fence[a] and explains why he thinks that map is misleading.

[a] https://stemlounge.com/content/images/2019/10/muddy_no_upper...

I've failed to find any text in the article that discusses why that map is misleading.

There's a section that discusses why it would be visually misleading to use a logarithmic scale, but I can't find anything that justifies the use of the statistical upper fence.

I'd be very interested if anyone could comment on why the statistical upper fence is considered less misleading for visualizing election results.

It’s kind of like the exposure on a photograph where there are some extremely bright spots, as bright as the sun. If you set the exposure to not clip those bright spots, you won’t be able to see the rest of the photo. Due to not being able to find a scale that faithfully reproduced the entire dynamic range (without visually equating population mountains to population plains), some clipping had to occur; a compromise. We could clip the few outliers using a standard statistical convention and retain the highs, mids, and lows that make up the rest of the image, or keep the statistical outliers and end up clipping pretty much everything else as a result. I included both renditions.

You’re correct that I didn’t say the overexposed graph is misleading. Neither graph is wrong, they both tell truths.

An article could probably be written about the one you’re inquiring about, that discusses how Los Angeles County, Cook County, and ~eight other counties are unlike any other county in the US in terms of population. Completely in a league of their own. As distant as they are, they may be more alike with each other than the counties surrounding them.

But there is also a lot of information to be gleaned from the photograph with the upper fence that leaves the non-outliers within gamut.

As the article describes, the 'artificial' ceiling is the standard quantity known in statistics as the 'upper fence':

https://en.wikipedia.org/wiki/Outlier#Tukey's_fences

I agree that it would be nice to have a scaled version of the top few counties - maybe an area-corrected inset of the most populous counties over the Pacific, Atlantic, Gulf, or Lake Michigan.

Would Cook county fit in the lake, or would it need to overlay Canada? It's half the population of the state of Michigan...

Regardless, hoping this gets updated for 2020.

Reply on news.ycombinator.com