STAT 135, FALL 06
Ani Adhikari
Week 1. Review of most-needed probability facts, in
the context of estimating the population mean (or proportion, in the
dichotomous case) from a random sample drawn with replacement. Confidence intervals.
Week 2. As above, when the sampling is without
replacement. Careful derivation of unbiased estimates of the
population mean/proportion, and population variance. Confidence
intervals, again. [Chapter 7 Sections 1-3.]
Week 3. Estimating
parameters in a distribution: the method of moments. [Chapter 8
Sections 1-4.] Review of how to compute with gamma densities, and the
special case of the chi-squared. Also, some basic technique: moment
generating functions.
[Chapter 4 Section 5.]
Week 4. Approximation by
the delta-method, or propagation of error. [Chapter 4 Section 6.]
This ends the discussion on method of moments. A more powerful
way of estimating parameters in a distribution: the method of maximum
likelihood, and good properties of the MLE when the sample is large. [Chaper 8 Section 5.]
Week 5. Proofs (sort of)
of the large sample properties of the MLE. [Chapter 8 Section 5.]
Efficiency [Chapter 8 Section 7] and sufficiency [Chapter 8
Section 8].
Week 6. The
language of testing, likelihood ratio tests, Neyman-Pearson Lemma
[Chapter 9 Sections
1-3]. Duality of tests and confidence intervals [Chapter 9 Section 3.]
Generalized likelihood ratio tests [Chapter 9 Section 4]. Calculating
power [lecture, homework 6].
Week 7. Parametric tests
for means [Chapter 9 Section 4, Chapter 11 Sections 1-2.2, 3.1].
This includes tests for proportions [lecture, homework]. Uniformly
most powerful tests [Chapter 9 Section 2.3]. Chi-squared test
(theory) for the multinomial [Chapter 9 Section 5].
Week 8. Chi-squared tests (method) for the multinomial [Chapter 9 Section 5 and Chapter 13 Sections 3-4]. Brief review and midterm. What you need to know for the midterm is listed here.
Week 9. Paired comparisons:
McNemar's test [Chapter 13 Section 5], Wilcoxon's signed-rank
test [Chapter 11 Sec 3.2]. Nonparametric tests for comparing two
independent samples: Wilcoxon's rank-sum test, also known as the
Mann-Whitney test [Chapter 11 Section 2.3].
Week 10. Comparing several
samples: Bonferroni's method [page 487, numbered 12.2.2.2!].
One-way ANOVA: decomposition of sum of squares, F-test for
equality of means (based on normality and other assumptions) [Chapter
12 Section 2]; Kruskal-Wallis nonparametric test for equality of distributions
[Chapter 12 Section 2.3].
Week 11. The least-squared predictor of Y given X [Chapter 4 Section 4.2]. The least-squares linear predictor of y given x
[Chapter 14 Section 1], and the regression effect [Chapter 14 Section
2.3]. The bivariate normal distribution (review of 134 material)
and inference in the simple linear regression [Chapter 14 Section 2, handout].
Week 12. Complete the
discussion of inference in simple linear regression [Chapter 14 Section
2, handout; which, along with the results of HW 9, includes most of the
answers to Problems 10-14 of Chapter 14]. Parallels and
differences between multiple regression and simple regression [Chapter
14 Section 4.5 (ignore the matrices for now) and Section 5].
Begin the mathematics of multiple regression [Chapter 14 Section
3].
Week 13. Inference in multiple
regression [Chapter 14 Section 4]. The proofs we did look
different from those in the text but they cover the same ground:
Theorems A and B of 14.4.1, Theorems A and B of 14.4.2, Theorem A of
14.4.3, Theorem A of 14.4.4, the t-test of 14.4.5. In addition we have the sum of squares decomposition and the F-test of whether all the slopes are 0.
Week 14. The partial F-test of whether a subset of the slopes is 0. Then on to Bayesian estimation [Chapter 8 Section 6 and Example E of 3.5.2].
Week 15. Review.