MrUnderbridge posted:

Wait, what's Islington then?

Dave Grool posted:

This is not consistent with my experience of the pac nw

Paladinus posted:

If it helps, it's supposed to be percentage.

jjack229 posted:

A scatter plot with each case as a single, separately colored dot and fatality rate and infection rate as the two axes would be interesting. Not sure if it would be all that useful.

DontMockMySmock posted:

Put a regression line on it

ultrafilter posted:

Chicken pox is highly leveraged and is driving the fit of your model. Use robust regression.

vyelkin posted:

Put both axes on log scales imo

taqueso posted:

but i need the trend line for my report

Karia posted:

Incubation is a color scale, obviously. But there's some important information that everyone else missed: size is based on when the disease emerged (as determined by a very brief wikipedia search.) Since I can't graph negative sizes to show BCE/CE, I decided the best way to do it is to normalize the size based on the beginning of recorded history, ~5000 years ago. I was too lazy to get it to write BCE, so enjoy the negative CE years values in the labels.

This very important information makes a startling fact obvious: we can see that the center part of the graph is occupied only by smallpox, which is a very old disease. Based on the fact that there are no large (i.e., new) dots in the center of the graph, it must be impossible for any new diseases to emerge that have both infection rate greater than 3 and fatality rate greater than 10%. The graph shows it, so it must be true. This is how science works!



EDIT:



More exciting science news! Not only did I fix the dates (it was rounding to the nearest 200 years because of an integer cast on the scaling), but when you add in a line chart in order of date, it shows a very clear straight-line path from MERS to COVID on the log-log chart. Much straighter than any historical period beforehand, at least when using this particular smoothing function. It's also very interesting that the rough log-Euclidean distance is about the same for every jump between diseases. That lets us predict what the next disease will be!

Based on this, I have a message for the CDC: in roughly 2022, my model expects a disease with an infection rate of roughly 20 and a fatality rate of >1%. Start preparing now!

taqueso posted:

thanks for the attempt but I needed a more accurate trendline

taqueso posted:

holy poo poo groundfloor

watching a nobel prize in the making

Tree Bucket posted:

Can we have a specially-jpeg'd version of this information for posting to Facebook?

T.C. posted:

Son of Thunderbeast posted:

Kangxi posted:

This is likely what is called a "dead cat bounce"

dwarf74 posted:

That's actually a pretty important finding of cognitive psychology, called the Stroop Effect.

C-Euro posted:

https://twitter.com/FiveThirtyEight/status/1298740505598210050

Legendary Ptarmigan posted:

Horizontal axis is 2016 Trump vote share, as a percentage for each media market. The seven different graphs each have their own axis from 20-80%. Vertical axis is sort of shared and represents percentage of searches pertaining to that sport compared to the total of all seven.

The biggest unintentional takeaway from their data is that the NFL is a juggernaut, being almost as popular as the other six activities combined and evenly popular across partisan lines.

Mr.Radar posted:

Hmm, yes, when I think of camping I think of food that needs to stay frozen before preparation.

Son of Thunderbeast posted:

This could make sense if the 0 on the X axis is between the two Es in "several times a week"

E: actually read the rest of the viz and never mind wow

FreeKillB posted:

Very helpful, thank you.

SgtScruffy posted:

That fuckin' bridge in Whodunnit that causes you to drop the matches