January 31, 2005
Basketball Data Source
In class, we will be using data on last season's UNC ACC basketball games. You will want to get a copy of this data and study its statistical properties because it might be on the final exam.
Last year, UNC played 16 ACC games. We played each of the other 8 schools twice, once at home and once away. We won 8 of these contests. The dataset has the scores for each of the 20 minute periods (2 per game).
Variable Definitions:
n = game number. game = opponent. at = (0 for home game, 1 for away game). nc1 / nc2 = points scored by UNC in the first and second halves. opp1 / opp2 = points scored by the opponent in the first and second halves.
Variables Created In Class:
generate nc = nc1 + nc2 (total UNC points)
generate opp = opp1 + opp2
generate win = nc > opp
generate diff1 = nc1 - opp1
generate diff2 = nc2 - opp2
generate diff = nc - opp
Sharp observers might wonder about defining "win" in terms of the score after two halves. In fact, three games were tied at that point and went into overtime. Unfortunately, UNC lost all three games, making the definition of "win" appropriate.
Posted by bparke at 11:24 PM | Comments (0)
January 27, 2005
Homework
Posted by bparke at 11:32 PM | Comments (0)
Hypothesis Tests
We talked about hypothesis testing in terms of a study of canine learning with a small sample and a alrge sample of Duke field goal attempts in their loss to Maryland.
Posted by bparke at 09:32 PM | Comments (0)
January 25, 2005
Confidence Intervals
The Central Limit Theorem is the basis for the most basic confidence intervals.
We can, for example, apply the CLT to a binomial distribution with a large number of observations.
The math:
It is important to understand that the confidence interval is stochastic, not the unknown true parameter.
Posted by bparke at 09:30 PM | Comments (0)
January 18, 2005
Probability Theory
Statistics and Probablity Theory alternative paths connecting the truth with the observed world. You study probability theory first for pretty much the same reason you study differentiation before you study integration.
(lectures from 1/18 and 1/20)
What is a probability?
The Law of Large Numbers is essentially the practical definition of probability. It implies that the sample mean converges to the population expected value.
The sample variance converges to the population variance.
The "aX+b" rules will prove invaluable.
We use the aX+b rules to calculate probabilities for normal distributions using the standard normal table.
This is similar in spirit to the translation between scales on a thermometer.
Posted by bparke at 09:30 PM | Comments (0)
January 11, 2005
Introduction
Posted by bparke at 09:29 PM | Comments (0)