Friday, 23 May 2014

Dismantling Data.

I spotted this meme on a friend's facebook wall this morning, I think it is a perfect example of the manipulation and over extrapolation of statistics to support our personal positions. This person is one of those people who seem to think the marijuana is a wonder substance with no adverse effects... she has even argued that it is an effective cancer cure. A claim that is supported by zero clinical evidence. I have nothing against the consumption of marijuana, and accept its helpfulness with some medical conditions. My problem is that its advocates seem unwilling to accept that there are adverse effects to its consumption. Increased rates of throat cancer and related mental health issues for example.

Anyway, that isn't really relevant in this case, here's the meme.



So the backstory. All you need to know is as of January 1 2014, marijuana has been legal to sell and consume in private.
All figures taken from Denver police figures, link to the pdf at the bottom of the post.
Murder rates in Denver are indeed lower in comparison to this time last year. The figure for last year is 17, this year its 8. That's a 53% drop so the meme is right. I have two problems with it:

1. As the murder rates are very low anyway its difficult to take any statistical significance from them. One could proudly boast of a "50% drop" if your murder rate had gone from 2 to 1. It just isn't significant.

2. This seems to be a case of falsely implying that two events are correlated so they must be causally linked as well. This is the fallacious thinking that led many to believe that the MMR vaccine causes autism. I'll demonstrate below why causation doesn't equal causation.

Let's take another crime from the data and compare the rates last year to this year:
Cases of assault in the same months last year: 871.
Cases of assault since the legalisation of marijuana: 1376

Now that's a rise of 58% with a better sample size, would it be fair to imply from this that marijuana increases aggression leading to more assaults? No of course not.

Take another example:

Drunken and disorderly represents a rise of a whopping 1425% from this time last year to this! Now that is a figure that a person against the legalisation of weed could take and really run with. After all isn't marijuana frequently touted as a safer alternative to alcohol? This could be manipulated to imply that alcohol and marijuana are a potent cocktail, and its consumption leads to further alcohol abuse, not less.

Is this down to the legalisation of marijuana? Very doubtful.


So you can see, I can cherry pick this data to support either side of the debate. The overall crime rates have actually risen from this point last year.

What we are actually seeing here is the normal fluctuation between data sets. Its important not to over extrapolate data and draw conclusions whatever he may personally wish to see.


Denver crime rates 2013/14 comparison. http://www.denvergov.org/Portals/720/documents/statistics/2014/XCitywide_Reported_Offenses_2014.pdf