Posted by Luke in A while back I used to teach statistics… then I was asked not to return (but that is another story). One of the key things we used to teach the students was that you should rely on correlation as a measure of real world relationships beyond the simple fact that it describes the possibility of a relationship. What the hell… I hear you say… this is the problem solving blog. Well, allow me to retort!
Causal flaws
Whenever you hear people talking about we know from research that standing in the sun too long causes skin cancer you can be certain that there is probably (and I say probably with a capital P) a very good chance that the two are related. Here’s the difference… there is a high PROBABILITY that they are related. They may not actually be related. Although I am not going to get sunburnt for science to find out! I believe there is a link, and herein lies the problem. I believe it. It’s effective for me to believe that because he helps me to not get sunburnt. As I live in Brisbane, this is a good thing.
However, when you take this same valid logic and begin to apply it to social problems you have a whole new issue. Let’s take my wife’s favourite hobby horse: the housing crisis. We have so far, had endless amounts of predictions on what was likely to happen based on research. We have heard dropping rents, falling prices, rising rents, rising prices. Every week some moron gets on television and tells us some new misinformation about the pricing of houses in Australia. Here’s the catch, they are all based on statistics and they are all based on the idea that someone thinks it relevant to spew them out as if they were the truth.
The problem with stats
Now statistics can show us interesting things. I will be the first to admit that but I would like to point out a very obvious if not overlooked fact: they are created by people who see relationships in the world. No people, no relationships. That data, in any case, comes from the human collective that imposes it’s thinking on top of it. We collect the data, we analyse it, we find the patterns, we report on them. These patterns are not real they are the interpretations of a scientific process… something the media in this country doesn’t tell us. So when we hear statistics we say “reality”. So here is a thought experiment I used to run for my ill-fated statistics students.
Cows and Gun Usage
Let’s say for argument sake that I wandered into a rural university (hey I am up for it) and measured the amount of cows in relationship to the amount of gun usage. I took my measurement stick (instrument) and postulated a theory. The increase in gun ownership can be linked to the increasing cow population. So I measured it and found a relationship that shows an 89% connection. Or in other words, I am 89% sure that the increase in gun ownership can be proven by the increasing cow population. Most people in the class say, “well that’s stupid!” and they are right. Because my pre-existing understandings of what cows are stops me from making the connections to gun usage. However, the data tells me there is a relationship. It is however spurious. It really doesn’t matter because it’s stupid.
What are they saying… not the data… the media
So now, we move back to my rant. The media. They collect all kinds of dodgy data and present them to the general populace as being facts. In the world of statistics we base our assumptions on the fact there could be at some stage a population big enough that would describe the relationship perfectly. But since, we can’t survey every cow, or every gun owner we take a best guess and base our assumptions on a sample. That sample then becomes very important.
Samples…
If I was trying to find the people who didn’t have a home, I would sample renters, the homeless, caravan park dwellers and the like. I wouldn’t go to the mansions in the inner city. I very much doubt John H. Fancypants would give me the data I was looking for. I would find those people who didn’t have a home and survey them. Or in my example above I would look for gun owning cow farmers. Hence, I have identified my sample group and predetermined my results to a large extent. And it’s this predetermined results set that interests me. It’s precisely that, a sample based upon an assumption I have made as a researcher. The measurement of my assumption can help me learn, there is no doubt, but all it does is prove that my assumption is wrong, or right. It does not prove the reality of anything. It merely gives me an insight into a possible relationship.
Rant OVER!
So what the hell am I saying? This, statistics is simply the science of finding out things we are interested in. It’s not a truth finder. It can never widely represent the whole picture, it is limited to it’s hypothesis, it’s testing mechanism and all of the other bits and pieces. It’s about drilling down from a theoretical assumption. So next time you hear a news presenter ramble on about “facts” ask yourself this: Who did they survey, what questions did they ask, where are they from? What you usually get is something like: 1000 shoppers were interviewed in Brisbane and 90% said they had plenty of money and aren’t worried about the economic crisis. The question is: who responded, what was there socio-economic background and what shop where they found in? If it’s somewhere ritzy… the data is crap. It’s lacks that fundamental element of any good piece of research… perspective. So remember… think for yourself and don’t believe the hype!
Powered by MightyAdsense
