Hacker News new | past | comments | ask | show | jobs | submit login

> How about instead of SAMPLING, and without performing any sort of "moving average", you just SHOW the homicide rate ...

The answer is simple -- it serves no purpose. If the point is to establish historical trends, and in particular if it's desirable to create a future trend line, the first thing to do is avoid the classic mistake of examining the raw data. (This is true in climate studies also, for the same reasons.)

In this case, a moving average meant to confirm or refute the thesis that crime rates are higher now than in the past, is best conducted with a long interval. Such an interval shows that crime per capita is much, much higher now than it was in the past, along with population (which also fluctuates over time).

The 1930s were a historical anomaly -- the Great Depression (an increased level of social upheaval and desperation), and prohibition coming to an end (which meant that gangsters had to think of a new way to make a living), arriving at once, make it unrepresentative. Any statistician hoping to establish social trends would treat that period as an obstacle to producing a meaningful trend line.

> I don't see much point in applying "a reasonable moving average"

Yes, I got that. And I won't be likely to successfully explain to you why it's needed.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: