“polls show that 34% of americans will vote for–”

what polls? whomst is being polled? i have never once in my life been polled. what is the sample size? what is the sample demographic? is it really 34% of americans or is it 34% of americans who answer random numbers on their landline??? poll this dick

As a statistician, these are EXACTLY the questions you should ask when interpreting a poll.

In fact, you shouldn’t HAVE to ask. A data scientist doing their JOB will provide that information UP FRONT. How the random sample was taken, who was polled, and what demographics were potentially left out of the poll.

If that information isn’t there, don’t trust it.

So can we talk about the absolutely stunning duplicity going on here?

…that’s cheating!

HOLY SHIT

That is some DELIBERATELY EVIL data representation. Where is the super villain that designed this fucking graph?

“Okay murders spiked by like well over 1000% but if we turn the graph THIS way, it’s like they went DOWN! HOORAYYYY!!!”

For anyone confused: the red represents the murder rate. Look at the numbers at the y axis. The graph is upside down because, normally, a graph starts at 0 and then goes up, but this one ends at 0.

As we put it in German: never trust statistics you haven’t manipulated yourself

this is fucking disgusting

There, I fixed it.

Holy fuck this is utterly vile.

This is why I believe everyone should have a stats class in highschool. People can moan about math all the way want but things like this affect not only your life but society itself. Being able to read graphs properly, understand statistics… It frees you from so much.

autumn and the spouses were talking about the odd perception of the miraculous fandom as an oversexualized place so i ran the numbers and. come now. let's pick up the pace.

average among the six non-miraculous fandoms* was 20.679% and a one-tailed significance test for the hypothesis that miraculous is less than the average had a p-value < .00001 which is. pretty conclusive.

*fandoms arranged by number of work and sampled by convenience

Bonus statistics: the Really Funny Graph Awards

As I've pointed out before, it's often hard to notice voter fraud— even in large quantities— if you can't see, not just how many votes came in, but when and in what patterns. Accusations of fraud don't track fraud, they track controversy; the most fraud often happens in polls that nobody particularly objects to, because nobody was paying attention.

Unless you have a graph in front of you. Fortunately, we do! So here's a brief review of the graphs that made it very, very clear to the mod team group chat that someone was playing silly buggers.

Round 1:

Davekat vs Zolu would almost even look natural, if it weren't for that enormous spike at the day 4 mark. But what a spike! And Akeshu vs Supercorp has those spikes in the middle, but the beginning stages of the graph look maybe fine... if you weren't watching for the first two days, and didn't get to see the progression:

Round 2:

This, on the other hand, couldn't be mistaken for natural by anyone. Look at this nonsense. The stairstep lines! The sharpness of the peaks! The sharpness of the dropoffs, which is how you can tell that this isn't just the poll being reblogged by large accounts, it is one person putting in truly insane amounts of effort! The fraud continuing long after Hualian had a significant lead, apparently just to make sure Buddie couldn't possibly launch a counteroffensive! Isn't it beautiful!

Round 4-5:

And once again, Hualian voters— or, well, some particular Hualian voter— goes nuts. Usually in 1-day polls, the votes come in fast enough that even with a graph it's hard to see if anything's gone wonky. Not so here; that bend in both graphs at around the same time, where I can only assume our frauder stopped for the night and went to bed, is a work of art.

Round 6:

No visible irregularities in the graphs (I assume they were just happy with getting to the semifinals?) but I did see this ask pop up:

I didn't see responses from anyone who took them up on it.

Real talk: This sort of thing is the reason I run poll brackets. This is proof that one person with insane dedication and a lot of time really can be the backbone of a fandom. This is, and I know this is melodramatic but I am being entirely sincere here, a chart of human passion.

Davekat, Akeshu, and especially Hualian— someone loves you very, very much.

now in HD!

To all my writers who have a tough time with smut terms and not knowing which ones to use, I have found the holy grail for us.

This reddit user took a poll of 3,500 people and went really in depth with asking their favorite terminology, along with actual pie charts on what the readers preferred to see in their smut.

Here's the direct link to the Google doc with all the info!

this conveys sample bias better than any statistics course

Heres the thing you gotta understand about statistics. 

“Increases your chances by 80%” does not mean “there is now an 80% chance”. 

If your chances were previously 10%, your chances are now 18%, not 90%. 

if your chances were roughly 1%, they’re now just slightly less than 2%. 

thats how that works. 

Wow I don’t understand math at all

‘if you have a baby after 35, the chance of deformities goes up by 100%’ is a line I hear alot.

It goes up from .5% to 1%

I think my brain just stopped working

100% is just another way of saying twice more likely. So 100% more basically means multiply the number you do have by 2.

Imagine how many woman are scared to have kids because of that statistic

This is why I took stats instead of calc. Because I don’t build engineer bridges in my everyday life but I sure do read studies that affect how I might live my life if I misinterpret them.

I’m terrible at numbers and math but I knew this and I really take it for granted. The average person definitely assumes, quite understandably, that “600% INCREASE!!!” must always mean a whole lot even if it literally only means that one of something is now six of something. Politicians probably take a shitload of advantage of this confusion.

just remember that increased BY and increased TO are very different things.

Oh god I didn’t even think about that whole other layer of confusion. Yeah if you’ve got 100 people and one of them is sick, that’s 1% of them who are sick, so if it “increased BY 100%” then that means now two people are sick. If it’s “increased TO 100%” then all 100 people are sick.

Reblogging again for that last addition.

Toastystats: F/F, F/M, & M/M on AO3

Chapter 3: Ratings & Smut

New chapter up! One of the main things I've found is that if we look at how likely each relationship category is to be explicit, F/F < F/M < M/M:

And (as @flourish put it on @fansplaining episode Femstats February) it's not just that F/F is less likely to be tagged Explicit -- it's more likely to be tagged General Audiences (the Teen and Mature and Not Rated ratings are similarly likely in all the categories).

(This isn't mainly due to word count differences, btw -- it turns out that this "explicitness gap" exists for fanworks of every length I investigated! See the chapter for more graphs/discussion.)

One thing I wondered about was how much of this was due to the source material -- e.g., I know there are some canon F/F cartoon pairings, and what if it turned out that a lot of people liked to write F/F cartoon fic that is rated GA? To try to factor out that kind of source material bias, I decided to also look at ratings for Original Work posted to AO3:

Let’s take some tumblr population sample statistics

I just want to know the age breakdown of this website so reblog please

Reblog please! I'd like to see this one reach the far ends of Tumblr. ^^

Please reblog this so that I can get a bigger sample size, but DO NOT include anything in the tags about the results, as that could influence other responses

this is the fucking funniest thing ive ever seen im in tears of laughter. (right axis) . this is a work of fucking art

As you all can see when I started to purchase $10 pocket knives at gas stations in late 2017 and continued to do so once a year every year the gap between my annual military expenditure and that of the U.S. quickly began to close.