Jake Vanderplas - Statistics for Hackers - PyCon 2016.mp4
Vložit
- čas přidán 16. 06. 2016
- Speaker: Jake Vanderplas
Statistics has the reputation of being difficult to understand, but using some simple Python skills it can be made much more intuitive. This talk will cover several sampling-based approaches to solving statistical problems, and show you that if you can write a for-loop, you can do statistics.
Slides can be found at: speakerdeck.com/pycon2016 and github.com/PyCon/2016-slides
He has the best book which is freely available online.
A really nice intuitive intro to resampling methods in the space of 30 minutes. So many times, bootstrapping is just bundled into data science education without any attempt to make clear the specifics of what is actually being done, what the purpose is, and why it makes sense or is reasonable to use it.
Highly recommended!
Fluff Miller should probably study statistics instead of "data science"
I wish any of my statistics teachers had even a fraction of his creativity or legitimate desire to impart knowledge. Bravo.
Another great talk from Jake! Thanks for suggesting the resources at 32:38
Oh my god - brilliant talk.
CSC 207 gang is here
Stfu Richard
Clever and creative use of examples from Dr. Seuss makes statistical concepts fun!
Having just taken an intro stats course (I passed!), I love this video. I like how he talks about "the right level of abstraction." Actually, I think the whole "right level of abstraction" thing applies not only to stats, but to a lot of math. I know Allen B. Downey has written some books on understanding statistics using python (Think Stats, Think Bayes, etc), and I think the same approach ought to be taken for other branches of mathematics, such as trigonometry and calculus.
This really hits home with me. Especially during grad schools when you are absolutely ingrained with the solutions and methods that by the time you finished, you totally forgot about the real questions you were solving. Good reminder indeed! Thanks!
anyone watching this cause arnold told you to?
That's me
Same
same
Same
it feels like a pretty random thing to be included in the assignment
Very clearly presented. Well done!
He’s such a great presenter.
Excellent delivery, wish other stats teachers had the same charm, and creativity; would be nice to see a causal inference for hackers.
Great! I learn a lot from the video~
Excellent talk~
I loved Jake, his talk, the presentation, examples and the Dr. Seuss theme. Still, even as he tries not to make hard statements... he certainly does. Finding out 'differences between statistics and data science' is a great open discussion. Are they really any different? There is evidently an intuitive explanation to t-statistics, chi-square distributions and degrees of freedom. These terms, didn't come out of the blue...these are just tools developed by humans who were trying to solve or describe real world situations (and these people certainly had problem solving minds).
I enjoyed this talk.
Fabulous - many thanks
Great talk
Quant_King's post on Hacker Statistics on reddit brought me here.
Awesome
The resampling book mentioned is out of print. Can somebody recommend an alternative?
The caveats at 21:00
20:44
14:38 This is anti-starbellied Sneech half-truth propaganda. You need a higher Sneech sample size. With your low sample size, it was almost guaranteed that the 6.6 point average difference would not be significant. Sample more Sneeches, and you may just find that average difference becomes quite significant at p =0.05
Coming here via doc py lib random see also documentation
When you know the analytical answer for something, using simulation is dumb. To give an example, we know the equation for the sum of the first n natural numbers. So we can calculate this in O(1). Now if you do not use the equation and just use for loop to calculate the sum, the cost is O(n). Why on earth would you go for O(n) algortihm when you have an O(1) algorithm? Another example would be using bogosort intead of quicksort.