nypd masters of statistics

[cyclops]

Breaking news:
"November 26th was the first day in 2012 with no murders in New York. Up to that days 366 murders occurred in 2012. The cops can't remember this happened before in history, a day without murder."
Cyclops says: "If you are such stupid you probably haven't solved a single murder yet. "

[Horibe]

Well, I do not mind looking stupid.... Murders are down in New York City. November 26th was the 331st day of the year. So, with 366 murders, in 330 days, it is certainly possible that this was the first murderless day of 2012. Sorry, I just do not know what you think is funny here.

[cyclops]

Because it is about as improbable as me beating a 9p. Unless there is a serial killer that commits a daily murder but had the flue last monday. Otherwise, most probably, there must have been plenty of murderless days already in 2012.

[Horibe]

Because it is about as improbable as me beating a 9p. Unless there is a serial killer that commits a daily murder but had the flue last monday. Otherwise, most probably, there must have been plenty of murderless days already in 2012.

Ok...that makes sense. It is statistically improbable.

However, I googled the story. In fact, what was remarkable for that day was not that there was no murder - it was a day without a violent crime of any kind, murder or otherwise. Quote below from an article:

There was not a single reported slaying, stabbing, shooting or knifing in any of the five boroughs on Monday, according to the New York Police Department.

"It is unusual in a city of 8 million people, but we never read that much into one day," said Deputy Police Commissioner Paul Browne, who said it was the "first time in memory" that the city had had such a lull in violent crime.

This is even more statistically improbable, but I am not sure Mr. Browne is stupid, assuming he has his facts, for this particular day, correct.

[cyclops]

Well, then my newspaper and our national television got it wrong. They claim to cite Reuters.
So I checked it at Reuters. And they reported as you stated. So maybe I am stupid for not having checked Reuters before or our national television and my ( mostly serious ) newspaper are.

[Bill Spight]

What I find remarkable is the 366 number. That sounds more like New York City 100 years ago to me.

[hyperpape]

One of the surprising facts that is rarely reported by the media, and never reported with adequate emphasis is that violent crime has rapidly decreased over the past twenty years (or thereabouts. I can't remember the exact time-frame).

[cyclops]

The Netherlands has a population of 16.6 million and a murder rate of 166 / year ( 2011 ).
Other recent casuality rates. traffic = 661/year, suicide = 1400/year, drugs = 94/year

[cyclops]

Today in the same dutch newspaper "De Volkskrant" ( supposed to be a quality newspaper ) the same stupid story again: "last monday the first murderless day in NY". And more nice, a photograph and a story about a NY policeman who buys a pair of shoes for a homeless. His name: Deprimo. It seems the NYPD got a new public relations officer recently. Only happy stories from now on.

[jts]

Kind of an embarrassment for Dutch journalism, no? Presumably this means one really awkward mistranslation plus multiple cases of plagiarism...

[p2501]

Am I the only one here wondering what cyclops is talking about?

[hibbs]

Am I the only one here wondering what cyclops is talking about?

If you assume that a murder is a random event, then the number of murders per day should follow a poisson distribution.
In this case, the count of murders per day is around 1.1

A simulation of the frequencies of a poisson distribution with a mean of 1.1 murders per day yields:
Days with 0 murders: 114
Days with 1 murder: 122
Days with 2 murders: 64
Days with 3 murders: 22
Days with 4 murders: 7
Days with 5 murders: 1
Days with 6 murders: 1

So really, you would expect that around one third of all days there should be no murder.
The probability of having really exactly around 1 murder per day and no days without murder for 311 days is so absurdly small that it really could only be caused by a serial killer who took one day off. And the police certainly should have noticed that there is a totally non-random pattern...
(Of course neglecting that the story was incorrectly reported in the newspaper. It might indeed be a very unlikely event of having one day without a single violent crime reported)

[mitsun]

This is even more statistically improbable ...

Actually, the whole point is that this scenario is statistically much more probable. While it is highly improbable to have a single day with no violent crimes, it is vastly more improbable to have a string of 300 days with at least one murder.

If the likelihood of zero murders on any random day is 1/3, the probability of 300 consecutive days with at least one murder is (2/3)^300 ~ 10^-53. This scenario is completely impossible, so the original report must be wrong.

If the likelihood of zero violent crimes on any random day is 1/1000 (just to pick a small number), the probability of 300 consecutive days with one or more violent crimes is (999/1000)^300 = 74%, and the probability of breaking this streak on the next day is 1/1000. This scenario is not impossible, and is perhaps rare enough to warrant a newspaper report.

[p2501]

Sorry, but isn't it obvious that what was meant was, that that day no murders were reported - meaning murders still may have occured unreported?

Way to go mocking the nypd, by taking every word literally in a quoted out of context sentence in a newspaper.

[topazg]

Man, us Go players are geeks

Papers have fun with statistics, sometimes with some level of reasonableness, sometimes unintentionally without, sometimes deliberately to spin things. Every organisation that is consulted by a paper may do the same, and then being misquoted gets added into the mix.

It's been a while since I read a stat in a newspaper and gave it any interest or credence (regardless of their accuracy).

C'est la vie!