ECON1203 FINAL EXAM TOPICS OVERVIEW Blair Wang BIT12DESCRIPTIVE STATISTICS GAMES & PROBABILITY Week 1 Week 2 Week 3 Week 4 Week 5 Week 6 • Frequency distributions and histograms • Shapes of distributions • Describing bivariate relations • Central tendency (location) • Dispersion (spread) • Association • Linear regression, incl. OLS (Ordinary Least Squares) • Data collection, incl. random sampling • Probability distributions incl. marginal, conditional, joint • Independence defined as P(Y|X) = P(Y) • Replacement when sampling • Random variables • Discrete prob. dist. • Mathematical expectation • Discrete probability dist. e.g. Binomial • 4 requirements for Binomial (n trials, Bernoulli, outcomes assigned prob., independent trials) • Meaning of iid. • Continuous probability dist. e.g. uniform pdf f(x) and Cdf F(x) • Normal dist. • Normal approx. to the Binomial • Estimators incl. properties INFERENCE FROM SAMPLES REGRESSION & CHI-SQUARED Week 7 Week 8 Week 9 Week 10 Week 11 Week 12 • Sampling error as param. minus statistic • SDOM (x-bar) • CLT states SDOM~N() for n>30 regardless of underlying dist. • Standard error (se) as sigma for SDOM • CI, incl. rv nature and half-width formula (Z_alpha/2)*(σ/sqrt(n)) • Introduction to confidence level as 1- alpha • Hypothesis Testing • Type I and II errors, incl. alpha and beta notation and conditional prob. notation • P-values • Power of test incl. twin x- bar dist's diagram • Student's t dist. for (underlying != normal && n < 30), incl. v=df=n-1 • Sampling dist. of sampling proportion P~N(p, pq/n) based on X~N(np, npq) • Equation components incl. (in)dependent (explanatory) variables, disturbances/errors • 7 assumptions of classical linear regression • Engel curve (income- expenditure rel'ship) • Variance decomposition SST=SSR+SSE (total = regression + error) and the derived R^2 • SEE (se of estimate) as estimator for sigma in Lin. Reg. • Inference and explanatory power • Dummy variable to test Boolean condition • Prediction in linear regression as Y = E(Y|X) + epsilon • Theta transformation • Interpretation of "partial regression coeffs" as ceteris-paribus marginal effect • Multiple regression model incl. A8 • "Linear" referring to coeffs, not variables • Hypothesis testing - Excel reports coeff=0 • Non-linear relationships e.g. "age and age^2" • Adjusted R^2 which considers n and #variables used • Good regression presentation reports SEE (se) instead of t-stat or P-value (Excel has all) • Chi-squared dist. as hypothesis test for population sigma^2 • Chi-squared for goodness of fit and contingency model, incl. df based on #cells, expanding % to n, and "multinomial" concept