Three Things

[sebenikela]

stultiloquentia gave me Three Things to talk about. If you would like three random things, leave a comment and I can come up with some Things.

1. Seagulls

Assholes, but in a fun way? I've never lived near enough to a large enough body of water to have significant interactions with them.

2. Astronaut food

So it turns out apparently Astronaut Ice Cream and suchlike, the stuff they sell in the Air and Space Museum Gift Shop, is actually not what actual astronauts eat. Probably this is for the best, because that stuff is fun as a random weird thing, but would not be fun as actual staple food.

There is an espresso machine on the International Space Station, courtesy of the Italians (of course) which is called "ISSpresso" because scientists love a bad pun. You have to drink out of a special zero-g cup that uses surface tension and capillary action instead of gravity.

3. Data for Progress

Capital-letters-version:

We are a multidisciplinary group of experts using state of the art data science techniques to support progressive activists and causes. Our current research areas include:

• Multilevel Regression and Poststratification (MRP) analysis to provide reliable sub-national opinion estimates on progressive issues.
• Deep learning textual analysis of media.
• Data mining and analyzing social media data for politicians and pundits to find interesting trends and patterns.

We distribute our research over the internet because data can only help interpret the world. The point is to change it.

I did not actually know about this organization, so thanks for the tip!

The above and their about page in general is buzzword-heavy enough to ping my bullshit warning sensors, but the actual work they are doing seems to be quite sensible.

Related to tech buzzwords, a friend of mine has a policy where he replaces "machine learning" with "linear regression", "artificial intelligence" with "numerical integration" and "blockchain" with "spreadsheet." This is an oversimplification, of course, but not an egregious one. Basically--if the thing still sounds worthwhile after replacing those words, it probably is! If it sounds stupid, then probably it's still stupid with the fancy words in.
/tangent

Because I can, I'm going to add a bit (eta: lot) on "data for progress" small-letters version:

Lots of people are very excited about using "Big Data" approaches in agriculture and int'l development spaces. also everywhere else but these are the two I can talk about intelligently. ish. There are two major problems with this trend/fad/thing:

1. There is not enough data
2. The data we have is not very reliable

This is true for crop production, national incomes, health outcomes... (There's a book I read on this called Poor Numbers which was quite interesting)

This is why if by some weird twist of fate I become extremely wealthy, I am going to start the Foundation For Boring Research.

Because nobody wants to fund "we trained national statistics offices in double-entry data entry and database management." They might be willing to fund "HEY LOOK TABLETS!" And for the record, tablets can be totally useful and appropriate for some applications, and they can make the process of data collection and analysis easier. But they don't solve the fundamental problem.

I mean, the fundamental problem is that there's no real incentive to get these numbers right and lots of incentives to get them wrong, but I don't know how to fix that one. A surprisingly (to internet denizens) prevalent additional problem is people who straight up do not know how Excel works, or how to do any statistics beyond a standard deviation. Some brilliant, persistent and clue-full people I know are working on the African Maths Initiative (which is obviously based in the UK where "maths" is plural). They do... all sorts of stuff around math/stats training from maths camps for kids to e-courses for University level statistics, to, yes, tablets and apps. And probably a bunch more stuff because I haven't talked to those guys in a while.

(Other things the Foundation for Boring Research would fund include medical/agricultural/economic/whatever trial replication, on-farm experimentation that is in no way novel other than "we're doing it with different people in a different place" and probably something about how to build roads that can be maintained, ish, by locals with shovels when the government ignores them.)

Anyway. More data analysis is great, but when people keep using the same shitty data that everyone KNOWS is shitty (looking at you FAO and yes I'm guilty of this too), then what's really the point?

Also, a lot of questions can't be answered with data. "How do you build a functional health system" has too many moving parts for any data to provide more than a small part of the answer.

tl;dr this is why I love data analysis and statistics but tend to be the one going "your data is flawed your methods are overly complicated and your results are irrelevant, please try including some qualitative data and talking to actual humans and try again."

Comments

[lilalanor]

I love your stats and data analysis thoughts especially <3