Interesting Links for 01-06-2015

[andrewducker]

• The Cheetah murderbot can jump now...
  • (tags: robots video )
• People are learning the wrong lessons from That Chocolate Study
  • (tags: fraud statistics research )
• Mad Max Fury Road Ponies
  • (tags: movies mylittlepony apocalypse )
• 1 in 4 women in the US is a victim of domestic violence, those numbers jump to 1 in 2 if they are married to a cop.
  • (tags:

Comments

[steer]

Nice article about the chocolate study. Yes some of the things in the chocolate study did draw my ire -- in particular the idea that you can never prove anything with a study with 16 participants. Much of modern statistics is founded on tests of that size and in specific the need to answer questions about samples of that size because experimentation is expensive.

[andrewducker]

Yeah, there's a lot of cargo cult science going on out there, with people repeating "Correlation is not causality" and "This research looks underpowered" as if they were magic spells to ward off ignorance - and without understanding the stuff behind it, and whether what they're saying makes sense in context.

If you have ten people, feed five of them a real pill and five of them a placebo, and the five that take the real pill all grow an extra foot in height, while the five that took the placebo continue to be continue being normal height, then that's a really strong result on a tiny sample size. Small sample sizes are rubbish for things when data is noisy, and results are feeble. If your results are strong, or data lacks any noise, then you can get by with a tiny sample.

(That's my understanding, anyway. I don't have the stats understanding to prove any of it, and I'm very happy to be corrected, because I know how ignorant I am!)

[steer]

Yes, pretty much exactly this -- your example is a compelling one. If you see a small sample size with a significant p-value it's because some huge effect happened. You are likely to get that happening (assuming you calculate your p-value properly and with honesty) only when the effect size is large.

Noise is complicated but basically, what we would mean by an effect size being large is the effect size in comparison to the noise -- you might almost think of the signal to noise ratio by analogy. So you intuitively said that growing an extra foot in height is "large" because people don't generally do that (the 'noise' in height measurements usually within 1cm).

[gonzo21]

"In this case Conservatives would have won 242 seats (-89), Labour 208 (-24), the SNP 30 (-26), the Lib Dems 47 (+39), Plaid Cymru 5 (+2), UKIP 80 (+79) and the Greens 20 (+19)."

Ye gods, what a healthier democracy we would have eh.

(Even if it meant we'd currently have a Tory/UKIP government of insanity.)

[andrewducker]

Exactly. I'd rather have a strong group of parties, people feeling represented, and frankly a chance to show people what those parties do when given some power so that they can make more informed decisions!

Plus, of course, the _next_ election we'd almost certainly have more parties, more choice, and different decisions made by both the parties and the voters!

[gonzo21]

Just thinking how wonderful it would be for me to feel as if my vote actually made a difference, and COUNTED! It would be marvellous.

I mean I can understand why the Tories and Labour are so afraid of it, because I think it does as you say, lead to a future of many more parties, and the erosion of the two big power blocs in British politics. But, yeah.

I read today that one of the founders of America said that 'Every generation needs a revolution'. Because these power structures that govern us have to be constantly reinvented to suit modern needs.

And when was our last revolution?

[coth]

https://www.youtube.com/watch?v=svMO5VXOBYw

Enjoy!

(Sex education in kindergarten, btw)

[andrewducker]

I'll watch from home :-)

[andrewducker]

I love that book! Have bought it for multiple friends with kids!

[coth]

Me too.

[andrewducker]

I didn't. Which was silly, in retrospect, because I already believe that most people who use drugs at that level are basically self-medicating to deal with trauma.

[alitheapipkin]

I knew our drugs policy was stupid but that's a really damning indictment of just how much it is failing some of the most vulnerable people in our society and perpetuating problems rather than mitigating them.

[andrewducker]

Yup, we're taking people who have nothing to live for, and are using drugs to cope with that, and criminalising them. It's awful.

[livejournal]

Hello! Your entry got to top-25 of the most popular entries in LiveJournal!
Learn more about LiveJournal Ratings in FAQ.

[skington]

There really isn't anyone posting on LiveJournal any more, is there?

[gonzo21]

I saw a post with about 6 comments get into the top 25 at the weekend there. Weekends are particularly dead on LJ.

[andrewducker]

392 views of that entry. Which isn't bad - I suspect that the vast majority of LJ posts are viewed by people in their own friend circle, and not elsewhere, so something with nearly 400 views is going to be at the top end.