Another exciting week, time to review those lessons. In honor of my own graduation, I’ll make it a stat-centric post.
1) Statistics is the core of modern science
Statistics tells you whether something is true, false, or merely anecdotal; it draws the line between a habit and random luck. It is the instrument that defines risk and its probabilistic consequences. Like all tools, however, its usefulness is measured by the skill of the practitioner. We unfortunately live in an age where it is very easy to fool people with statistics, because very few people understand the subject. You either have the mathematicians who leave it to a politician to translate the results into plain language or you have an analyst who can’t do the math so they just punch in numbers until they get a result that looks good.
2) The primary question of risk is consequence
It seems obvious on an intuitive level, but people often make mistakes applying this to the real world (i.e. modern finance). When you take risks, the only thought should be of the consequences. In finance, this is called leverage. In some situations, a single error can blow up in your face if you are improperly leveraged. An example is putting all of your money in one stock. In others, you can be totally wrong and face no consequences at all. An example is paying too much for parking.
3) Beware regression. Especially multiple regression
Science has fallen in love with correlation. It’s easy to do, easy to understand, and draws a single convenient line through a very messy cloud of data. The problem is that even very close relationships that you can find in research rarely if ever replicates itself in real life. This is especially true if you’re dealing with the behavior of human beings, such as in psychology, economics, or politics. The other problem is that correlation is a single number that tells you only that you have two relationships that seem to move together. Going from that single number to a policy recommendation that costs millions of dollars and affects many people’s lives (through treatment, policy, etc.) is a leap over a very large abyss. Your correlation number tells you nothing about causation, randomness, or hidden risk.
Just remember, you can find correlations between anything if you put enough variables in. That doesn’t tell you anything meaningful.
4) Know what you don’t know
The biggest problem with students and the ignorant is that they roll their eyes when you tell them about the unknown. They think that if you’re not giving them information, then they are learning nothing. People who know what they’re doing value this immensely. Every other episode of House highlights this.
5) Watch out for charlatans and survival bias
Survival bias is one of the most pervasive and dangerous consequences of statistics. This is how you get trite life tips like “all you have to do is work hard” or “volunteering looks good on your applications”. Most people with “how to” tips are charlatans and owe just as much of their success to luck as they do to skill or knowledge. Thousands of people try every year to follow the advice of someone on how to make their first million by the age of 24 and almost all of them fail (and the ones who succeed usually didn’t follow their advice at all). This is survival bias, that we follow the advice of survivors and assume that their way must be right because they made it. What you don’t see in bookstores is business books with titles like “How I failed in business and in life”. Similarly, academics publish results as though the experiment should run like silk. You never get a “this is how I wasted the last ten years of my life and got nothing” paper, even though it would be vastly more informational and truthful.
Goals from the week:
1) Place was cleaned enough for Yoko and Patrick
2) Ran 5 miles this week, far short of the 15 mile goal
3) Graduated and it was sweet
4) Sent out 22 resumes
Verdict: Getting lazy on the exercise. Fix it.