Let's talk a little about statistical independence. Loosely speaking, we say two events, A and B, are independent, if knowing that event A occurs does not change the probability that event B occurs.
E.g. Let A = the event that I play Rock Band 2 tomorrow. Let B = the event that Fast Eddie eats a salad for lunch tomorrow. These events are logically independent since my video gaming habits are not going to affect Eddie's lunching habits.
A little more complicated: Suppose you have a standard deck of playing cards. Suppose you're going to draw two cards. Let A be the even that the first card drawn is black. Let B be the event that the second card drawn is black.
Are these events independent?
Well, the probability that A occurs, denoted P(A), is 26/52 = # of black cards / # of cards total = 1/2
What's P(B)?
Well, if you drew a red card on the first draw, then P(B) = 26/51 (26 black cards but only 51 cards left)
If you drew a black card on the first draw, the P(B) = 25/51 (25 black cards (since you got one on the first draw) out of 51 cards left)
So P(B) changed DEPENDING on what you got on the first draw. So P(B) DEPENDS on what happened with A. Hence, these two events are dependent.
So why is this sometimes counter-intuitive? Suppose that I have a fair coin. By fair I mean that on each flip, the coin is as likely to land heads as it is to land tails. P(heads) = .5 = P(tails). Coin flips are independent. If I flip a coin and it lands heads then it's no more or less likely to come down heads on the next flip. Easy to say, but in practice this throws a lot of people off. Let's say I plan to flip a coin 10 times. The first 9 flips all come down heads
H H H H H H H H H
I pause for a moment and say, "what do you think it will land on the next, and last, flip?" We all (including me) have the urge to say "Tails! For the love of Buddha this string of heads is improbable!" This sentiment was dangerously echoed by some gamblers I observed when working at Treasure Island Resort and Casino. These gamblers would sit by the electronic roulette machine with notepads and they would write down the color and number of each outcome of the spin of the virtual roulette wheel. They would then attempt to use this information to help them predict the next outcome. At the most basic level of statistics just discussed, this would be a good strategy if the events were....drum roll please....dependent. If the spins of a roulette wheel were dependent then knowing something about previous outcomes could give you some information about future outcomes.
But...just like a coin flip, the spins of roulette wheel are INdependent. If that weren't true, then you would have to believe that the little wooden ball (or computerized wooden ball) somehow remembers what it has landed on and in the future will act upon that knowledge. If these men saw the ball had landed red 5 times in a row, they would probably bet on black for the next spin. This is FOLLY! Just because I flip 9 heads in a row doesn't mean the last flip is more likely to be tails. The coin doesn't remember that it just landed heads up 9 times in a row.
People have a false intuition about this empirical law called the Law of Averages, which, roughly put, states that thing even out after a while. I might get more into that in a different post, as it deserves it's own clever title.
Some real world applications? If you win the lottery using your favorite 10 digit number, there's no reason why you shouldn't keep playing that number afterwards.
If you haven't already, please, for me, watch (or read) Tom Stoppard's Rosencrantz and Guildenstern are Dead. It's the funniest, cleverest thing you'll ever put your eyes on. The first ten minutes are a beautiful homage to probability theory.
p.s. something to think about. suppose that people flip coins for fun. you know, for choosing between two movies, dates, desserts, colors of ties, etc. Suppose that the average coin is flipped 100 times in it's lifetime. Meaning someone flips the coin then buys something with it and the next person to get that coin flips it for some reason or another etc. Now, do you think that somewhere out there is a coin that's always landed heads? It's been passed around from person to person and no one single individual would notice, but could there be a coin out there that's been flipped 100 times and every time the coin has come down heads? How many coins would there need to be in circulation for this to have a better than 50% chance of happening? Is this coin magical?