If you'd like to support these videos:
[ Ссылка ]

Have you ever wondered how people say the word UHMMM when they talk? No? Uhhh... Well .... not much I can do about that now. Maybe check it out and you'll still find something interesting? Hope you enjoy and if you do, consider liking and/or subscribing! It means so much for the growth of the channel.

I am trying to aim for quality over quantity with these videos. If you want to support the channel consider checking out my patreon: patreon.com/NotDavid


#maths #stem

Chapters:
0:00 I Need a Real Hobby
1:38 Understanding The Data
2:44 A Four-Wheeled Vehicle of Transportation Analogy
4:22 Making a Graph
5:07 Why is the Graph so Fishy
6:34 Pop-Quiz for Nerds
6:45 Why is the Graph so Fishy
9:27 Finally the Results
11:28 Matt and Tom
13:59 Objection!


Thank you to all the people that allowed me to mention their name in the video, thank you to all the people doing public presentations, the Royal Institute for hosting so many talks and posting them online, and thank you to all the "participants".

Made using Blender
Credits At the End of the Video

Music:
Chris Doerksen - RPG store
Bandcamp: [ Ссылка ]

Lifeformed - 9-bit Expedition
Bandcamp: [ Ссылка ]

Not David - Start of a New Day

Toby Fox - Hotel

Chris Doerksen - Breather
Bandcamp: [ Ссылка ]

Lifeformed - Light Pollution
Bandcamp: [ Ссылка ]

JoJo4 - Great Days (instrumental)

Data for project: [ Ссылка ]

Notes:

1. 2:16 The data set has 40 uhmmm lists, though a couple come from repeat individuals. This was to test if people uhmmmed consistently. This does appear to be the case but I didn't have enough people to conclusively say so, so I didn't mention it. Moreover, as I discuss later in the video, not all of the people are science educators - for example at least two are politicians, which was to test if training reduced uhmming.

2. 8:31 Surrogate analysis involves creating new data that follows a known distribution. This can be difficult in general, but in Poisson this is really easy - lets say the rate of your real data is 1 event in 10 time units, then you make a list of length say 10000, where 1000 of those are `1' and the rest are `0', and then you randomly permute that list and it will be Poisson. Next you compare the distance between the real data and the surrogate data. One common approach is using the Kolmogorov-Smirnov distance, which measures the maximum difference between the cumulative distribution functions. If the distance is small, it suggests that the real data follows the assumed distribution (e.g., Poisson), while a large distance indicates otherwise.

3. 11:56 This is related to the previous note. Here what I mean to say is that Matt and Tom's Kolmogorov-Smirnov distances are large, so they are not likely to be Poisson. Pretty much everyone bellow Matt and Tom have small KS distances and so they are statistically likely to be Poisson. This is true even if that person is not on the line, and suggests that if we took more data they would approach the line.

Videos featured in my video:
Tom Scott: [ Ссылка ]
Matt Parker: [ Ссылка ]
Grant Sanderson: [ Ссылка ]

свернуть