[anamax]

> Above the rate for the year 1960, but not the rate for the decade 1960's.

My point is that the rates were changing so much during the 60s and 70s that speaking of averages loses too much information. 1960 and 1969 were very different as far as violence in the US is concerned.

> 5.1 vs. 5.4 for such a rare and random event is not particularly meaningful.

It's frequent enough to be statistically significant. We're not talking about the difference between 1 and 2 incidents in a population of 1M, we're talking about 10k incidents in a population of 180M.

The advantage of talking about murder (which includes manslaughter) as opposed to rape and drunk driving is that reporting isn't a big problem - the only argument is over the circumstance of death.

[Retric]

Do you really think the probability of someone that caused a fatal accident with a low BAC being charged with manslaughter was the same in 1960 and 2005?

As to being significant looking at different crime rates you tend to see similar trends but the "peak" year is often different. You can try to interpret it to mean something, but the reality is your looking at an imprecise estimate of an imprecise estimates of a highly random event it's really vary random after the first digit.

  1991 Forcible rape 106,593 
  1992 Forcible rape 109,062
  1991 Robbery 687,732
  1992 Robbery 672,478

Now look at which crimes rates up vs down from 1992 to 1993 etc.

[anamax]

> but the reality is your looking at an imprecise estimate of an imprecise estimates of a highly random event it's really vary random after the first digit.

Your assertion would be more convincing if your anecdote was had variance in the second position. Instead, it has variance in the third position (1065 and 1090) and less than half a unit in the second position (6877 and 6725)

The murder/manslaughter numbers are varying by 4-6x as much. (5.1 to 5.3/5.4)

[Retric]

My point was, you can ask questions like: Which was the peek year for crime? Why did rape not follow that trend? etc etc.

  Downward trend: - 2.2%
  1991 Robbery 687,732 
  1992 Robbery 672,478

  Upward trend: + 2.3%
  1991 Forcible rape 106,593
  1992 Forcible rape 109,062

Edit: By second digit I was talking about % change, (1065 and 1090) = 2.3% change as is 90 and 92.07. Clearly saying 98 to 99 is a change in the second digit where 100 and 109 is a change in the third is missing something.

However, X + 2% vs Y - 2% does not necessarily mean anything. If you want to understand people you need to look beyond any single year to find cause and effect.

PS: A heatwave can significantly change a city's crime rate, but on it's own there is no meaningful long term effect. Average things out over months and years and you can start to see meaning emerge from chaos.

[anamax]

> My point was, you can ask questions like: Which was the peek year for crime? Why did rape not follow that trend?

"What was the peak year for crime" assumes a definition of crime, a way to aggregate rape, murder, etc. While we can agree on one, that agreement isn't binding on others. That's why I didn't claim that there was a peak crime year.

That said, averaging the 60s doesn't make sense as there was too much change.

Demographics alone tell us that different kinds of crime will peak a different times.

> By second digit I was talking about % change,

Since I was talking about to 5.1 to 5.4, the 98/99 comparison isn't especially relevant.

> A heatwave can significantly change a city's crime rate, but on it's own there is no meaningful long term effect.

You're assuming that weather doesn't have trends or long term effects. I'd argue that the dust bowl did. Urban heat-islands cause persistent heatwaves. (Yes, that's artificial climate, but ....)