Methods Requirement

Biostatistics PhD students have requirements (prelims) in each of Theory, Methods and Applied Statistics/Data Analysis. Statistics PhD students need to pass two prelims and take one additional core sequence. The core sequences are:  theory; methods; stochastic processes; and computing. This proposal describes the Regression Methods requirement and associated PhD qualifying exam.

The Regression Methods core sequence is defined as the three courses 570, 571, and 572.

Biostatistics PhD students and Statistics PhD students that take Regression Methods as a core sequence are required to:

In addition, all Biostat PhD students, and Stat students taking the Methods prelim must:

Students who do not attain 3.0 in both 570 and 571 will have failed the prelim and will have the option of retaking these classes once more. In recent years there have been take-home midterm and final exams in 570 and 571; all students will be required to turn in two copies of these exams. Those exams that are failures, or close to failures, will be second marked (usually by the other 570/571 lecturer).

The Methods prelim presentations will occur in mid-June and will be attended by the team-teaching group, and any other faculty members who wish to attend. The prelim exam assessment will be based on input from paper advisors, and the exam committee.

The previous 572 course (Non-parametric regression) has been assigned a new number (527). This class is now targeted at the Masters level, and so may be taken by Stat and Biostat students in their first year (or later), after students have completed an introductory regression class.

A list of papers and advisors will be available (see examples below), but the student/advisor may also suggest their own paper. Of course, if the paper is chosen by the student the agreement of the advisor will be required. Suitability of all papers will be assessed by the Methods Requirement Committee, which will consist of 4 members, two from Stat and 2 from Biostat. Obvious candidate members are the lecturers of 570 and 571. The committee will have the final say on the suitability of the paper. The grading of the paper/oral will be carried out by the exam committee. The grading of the paper/oral will be carried out by the exam committee which will attend all talks.

In the written report and oral the students will be expected to:

  1. Summarize the main contributions and novelty of the paper. This may include reading additional literature in order to understand the paper’s genesis, and to assess the paper’s subsequent impact.
  2. Understand all of the analytical work in the paper, including ‘filling in the details’ of analytic arguments.
  3. Reproduce at least a subset of the simulation studies and data analyses in the paper.
  4. Critique the paper.

The student and faculty advisor will meet regularly (weekly) over the course of the paper study. These meetings may commence before Spring quarter begins. During Spring quarter, students taking 572 will meet as a group to discuss and present their papers.

The Methods prelim presents an opportunity for students and their faculty advisors to work with each other, on work related to substantial advances in methodology. The Methods prelim may (ideally) serve as a lead-in to thesis work.

Advantages:

Disadvantages:

Examples of potential papers:

1.      Liang KY, Zeger SL and Qaqish B (1992). Multivariate regression analyses for categorical data (with discussion). Journal of the Royal Statistical Society, Series B, 54, 3-40.

2.      Prentice and Sheppard (1995). Aggregate data studies of disease risk factors. Biometrika, 82, 113-125.

3.      Zeger SL, Liang K-Y and Albert PS (1988).  Models for longitudinal data: A generalized estimating equation approach, Biometrics, 44, 1049-1060.

4.      Cai T, Tian L, Solomon SD and Wei LJ (2008). Predicting future responses based on possibly mis-specified working models. Biometrika 95, 75-92.

5.      Gong, G. (1981). Pseudomaximum likelihood estimation: theory and applications. Ann. Statist. 9, 861-869.

6.      Zhang M, Tsiatis AA, Davidian M (2008). Improving efficiency of inferences in randomized clinical trials using auxiliary covariates. Biometrics, 64(3): 707-715.

7.      Gui, and Li (2005). Penalized Cox regression analysis in the high-dimensional and low-sample size settings, with applications to microarray gene expression data. Bioinformatics Vol. 21, 3001-3008.

8.      Donnelly CA, Laird NM, Ware JH (1995). Prediction and creation of smooth curves for temporally correlated longitudinal data.  JASA 90: 984-989.

The “Breakthroughs in Statistics” volumes will also provide papers. Papers may be recycled year-to-year, but a copy of all reports will be kept, to help protect against plagiarism.