Lecture 6

Paper 1 Select an area of human endeavor (e.g., a particular sport, marketing, politics, car manufacturing …) and write a 5 page paper on how statistics influenced it. Due on Thursday 9/29. Make sure you clearly reference your sources (including websites)

Probability: the study of randomness Randomness Basic Probability models – language Simulations

Randomness Which is random? A lot more “structure than people give credit”

The language of probability In statistics, random means more than just unpredictable or haphazard. A random phenomenon is a situation in which the outcome is uncertain, but there would be a definite distribution of outcomes if the situation were repeated many times under identical conditions. (And the same distribution would result if it were repeated many times again.)

The language of probability Examples Toss a coin, note whether it comes up H or T Take a SRS from a population, find proportion who call themselves Unaffiliated Choose 40 days at random from the 365 days of the year, check whether any day is repeated Not a random phenomenon: UNC-Duke game, … In each example but the last, we can ask: what would we expect for the distribution of outcomes after many repetitions? (That’s what makes them random phenomena.)

Probability is expected long-run proportion Results of a simulated series of 10,000 tosses of a fair coin tosses results so far expected results so far discrepancy (observed – expected) proportion of heads 1-10 8 H, 2 T 5 H, 5 T +3 H .8 11-50 23 H, 27 T 25 H, 25 T – 2 H .46 51-100 49 H, 51 T 50 H, 50 T – 1 H .49 101-500 248 H, 252 T 250 H, 250 T .466 501-1000 475 H, 525 T 500 H, 500 T – 25 H .4750 1001-10000 4952 H, 5048 T 5000 H, 5000 T – 48 H .4952 The “law of averages” applies to the proportion of heads, not the numbers of heads and tails. It doesn’t operate by making up a discrepancy, but by swamping it.

R code n=10000 p=.5 set.seed(2013) u=runif(n) x=(u>p) y=cumsum(x) print(y[c(10,50,100,500,1000,10000)]) print(c(10,50,100,500,1000,10000)-y[c(10,50,100,500,1000,10000)]) print(y[c(10,50,100,500,1000,10000)]-c(10,50,100,500,1000,10000)/2) print(y[c(10,50,100,500,1000,10000)]/c(10,50,100,500,1000,10000)) #PLOT THE RESULTS par(mfrow=c(2,2)) plot(1:30,x[1:30]) plot(1:n,y,type='l') plot(1:n,2*y-1:n,type='l') lines(c(0,n),c(0,0),lty=2) plot(1:n,y/1:n,type='l') lines(c(0,n),c(p,p),lty=2)

Plotting in R Commands plot(x,y,type=“l”) lines(x,y) points(x,y) Creates a new plot (without the type=“l” plots dots) lines(x,y) points(x,y) Adds additional plot without erasing the old one

Probability is expected long-run proportion The “law of averages” applies to the proportion of heads, not the numbers of heads and tails. It doesn’t operate by making up a discrepancy, but by swamping it. Illustration to follow: a simulated coin-tossing game. On each toss, you win $1 if the coin lands heads, and you lose $1 if it lands tails. So, after a number of tosses, your net gain is number of heads – number of tails. (Net loss if this is negative.) Should this be near zero after many tosses? Not necessarily. The proportions of heads and tails will be near .5, but that doesn’t mean the difference between numbers of heads and tails will be near zero. Use cointossdemo2

Subjective probability The probability based on repeated sequence is objective – independent on a user. Probability that UNC will win ACC title. Not objective – only one observation. We all have our own beliefs about that event. The probability can be evaluated using bets. If I am willing to bet 1:9 on UNC (win $9 if UNC wins, loose $1 if they lose) my probability is >10%, If I am willing to offer 1:9 bet on UNC (loose $9 if UNC wins, win $1 if they lose) – my probability is <10% Can be objectified for an individual using a series of bets. The fact that people play lottery/casino is a proof that we are poor judges of subjective probabilities. This is especially true for small probability events.

Betting sights Trade on future outcome – at expiry 100 if happens/ 0 if it does not http://www.paddypower.com/bet/politics/ http://tippie.uiowa.edu/iem/markets/ In the past – intrade.com (closed by regulators)