3 ms·
>In other words, the data is correlational, and not based on random assignment to an experimental condition and control condition. If you would like to see stu
by yetanotherphd 13y ago
>In other words, the data is correlational, and not based on random assignment to an experimental condition and control condition.
If you would like to see studies that use random or quasi-random selection into education, there is a huge literature in labor economics on this topic. For various reasons such literature would never make it onto HN.
See a very brief summary here
http://www.stlouisfed.org/publications/re/articles/?id=1866 http://www.stlouisfed.org/publications/re/articles/?id=1866
- sanxiyn 13y agoNow I am curious. Why would such literature never make it onto HN? (And eh... didn't you just make it onto HN?!) Thank you for the link anyway.
- yetanotherphd 13y agoBy "make it onto HN" I mean as a story - the comments tend to be much more varied. Mainstream economic research covers very similar topics, and uses very similar methods, to stories that make it onto HN. However since it isn't sold as "using data in cool ways", it's not immediately apparent that it is HN material. For example, a very famous economics paper uses school district boundaries to measure some effect (can't remember exactly what they measure). The idea is that people on opposite sides of boundaries should on average be very similar except for being in different districts. This hypothesis can be checked by looking for measurable jumps across boundaries. Another thing is that economics papers tend to use regression methods that look quite boring and don't involve interesting scatter plots. So to the untrained eye a blog post with scatter diagrams and maps can look like a novel use of data and technology, while in fact the methods are very standard.