SYLLABUS

SURV 742, SURVMETH 618

Inference from Complex Surveys
Winter 2013

Instructors: Robert Fay, Westat

BobFay@westat.com
Telephone: 240-314-2318 Fax: 301-294-2034

Brady T. West, University of Michigan-Ann Arbor (MPSM)

bwest@umich.edu
Telephone: 734-647-4615 Cell: 734-223-9793 Fax: 734-647-2440

Days: January 8 – April 16, 2013

March 5, University of Michigan break, but class will be held
March 19 No class, UMD break
14 sessions

Time: Tuesday afternoon, 3:00pm – 5:40pm

(10 minute break at approx. 4:15)

Places:
JPSM MPSM BLS
1218 Lefrak Hall Conference Perry G300 Postal Square
Room Building
University of Maryland University of Michigan Conference Center
College Park Ann Arbor

Texts:

Optional

Introduction to Variance Estimation (2007), 2nd edition by Kirk M. Wolter. Springer-

Verlag.

Applied Survey Data Analysis (2010), by Steven G. Heeringa, Brady T. West, and
Patricia A. Berglund. Chapman Hall-CRC Press.

Data Analysis using Stata (2012), 3rd Edition, by Ulrich Kohler and Frauke Kreuter. Stata
Press.

1
Classics Often Cited in the Notes:

Sampling Techniques (1977), 3rd edition by William G. Cochran. Wiley.

Model Assisted Survey Sampling (1992), by Carl-Erik Särndal, Bengt Swensson, and Jan
Hakan Wretman. Springer-Verlag.

Other Recommended References

Practical Methods for Design and Analysis of Complex Surveys (2004), 2nd edition by
Risto Lehtonen and Erkki J. Pahkinen. Wiley.

Sampling Statistics (2009) by Wayne A. Fuller. Wiley.

Analysis of Complex Surveys (1989) edited by C.J. Skinner, D. Holt, and T.M.F. Smith.
Wiley.

Analysis of Survey Data (2003) edited by R. Chambers and C.J. Skinner. Wiley.

Analysis of Health Surveys (1999) by E.L. Korn and B.I. Graubard. Wiley.

Complex Surveys: A Guide to Analysis using R (2010) by Thomas S. Lumley. Wiley.

Class website:
https://ctools.umich.edu/portal (C-Tools)
Class “chapters” (detailed lecture notes) and data sets will be posted here.
Please print the chapters each week before coming to class.
Separate instructions will be distributed on how to establish a “Friend” account at
the University of Michigan that will allow you to use C-Tools.

JPSM media site: Lectures will be recorded.

Go to http://jpsm.umd.edu/ and click on “Online Media” under “Links”. Click on
the Login link in the upper right-hand corner.
username: umd
password: jpsm

Readings:

PDF files of readings are on the class website under the link for Site/Info/Coursepack.

2
Software:

WesVar v.5.1—Free download from Westat.

http://www.westat.com/expertise/information_systems/WesVar/index.cfm
Click the “Documentation” link to download the WesVar manual.

SUDAAN 10.0—available on the JPSM network and the Michigan network to

MPSM students; not available to non-MPSM Michigan students. At
MPSM, SUDAAN is only available on the Unix server.

Stata 12.1—on the JPSM and Michigan networks. At Michigan, Stata 12.1 is on
both the Windows network and the Unix server.

R—Free download from http://www.r-project.org/CRAN. First, install R on your

computer by selecting a CRAN mirror site (preferably in the U.S.: IA
seems more stable than MD) and follow the installation menu. Then, open
R, from the “Packages” pull-down menu select “Set CRAN mirror,” which
can be the same site you used for R, then in the same menu select “Install
packages” and find the survey package in the alphabetically arranged
list. Thomas Lumley’s book (listed above) is highly recommended if you
choose to use R for this course.

SAS—SAS 9.2 and later versions support most calculations required for the
course, so students working in institutions with a strong SAS orientation
may consider this option. The JPSM computer lab is expected to have 9.3
by the start of class.

Advanced special students at UMD can use the JPSM computer lab. Contact
Duane Gilbert (dgilbert@survey.umd.edu) about access. There are four
computers in the lab, and he can also tell you about additional computers
elsewhere in LeFrak that you may be able to use. If you want to use the
lab outside of regular office hours, you will access to 1218 LeFrak
using your UMDID. Contact the JPSM coordinator, Stacy Hall
(shall@survey.umd.edu, 301-314-7911) to get access.

ISR network—To get access to the Unix server, you must fill out an Account
Request Form. Check with the administrative staff in MPSM to do this.
Documentation for the ISR Unix server is at
http://src.isr.umich.edu/unix/howto.html. These how-to guides cover
mapping network drives, using the Secure Shell Client, and other topics. If
there is strong interest, an orientation session can be arranged.

Office Hours: By appointment. Possible trial use of a chat room in CTools (TBA).
Your specific questions will usually be most quickly answered by email.

Teaching assistant: None.

3
Homework:
Assigned weekly, and due the following week. Answers can be emailed (PDF
format preferred), faxed to JPSM (301-314-7912), or paper copies can be put in
either Fay’s JPSM mailbox or West’s MPSM mailbox.

Grading:
Homework Assignments (65%)
Late assignments will be penalized.
1-2 days late: 10% off (Wed., Th. for homework due on Tuesday)
3-6 days late: 25% off (Fri.-Mon. for homework due on Tuesday)
7-10 days late: 50% off (Tue.-Th. for homework due on previous Tuesday)
11+ days late: no credit.

Take-home final exam (30%): distributed April 16, 2013, due at 5:00pm April 23.

Class participation (5%).

Course Topics and Tentative Lecture Dates

Class: 1
Day: January 8, 2013
Lecturer: West
Topic: Chapter 1. Introduction

- Syllabus Review
- Need for special methods for inference from survey data
- Types of sample designs, and impacts they have on relationships
among observed data
- Analysis of survey data
• Survey weights
• Weighting adjustments
• Impact on inference

Class: 2
Day: Jan 15, 2013
Lecturer: Fay
Topic: Chapter 2. Estimation of variance
- Reasons for estimating sampling variance for analysis
- Criteria for evaluating variance estimators
- Degrees of freedom of variance estimators
- Sampling error estimation for descriptive vs. analytic statistics

4
 Chapter 3. Techniques for Estimating Sampling Errors for Complex
Estimators
- Linear estimators vs. complex estimators
- Linearization vs. replication
- Use of transformations for deriving confidence intervals

Class: 3
Day: Jan 22, 2013
Lecturer: West
Topic: Chapter 4. Taylor Series Linearization, sections 4.1 to 4.6

- Technique of linearization (method and theory)

- Examples of linearization in practice
- Degrees of Freedom approximation

Class: 4
Day: Jan 29, 2013
Lecturer: West
Topic: Chapter 4. Taylor Series Linearization, sections 4.7 to 4.11

- Linearization for implicitly defined estimators

- Woodruff / Francisco-Fuller methods for quantiles.
- Computer software for obtaining sampling errors via linearization
(SUDAAN, Stata, R, etc.)

Class: 5
Day: Feb 5, 2013
Lecturer: Fay
Topic: Chapter 5. Balanced Repeated Replication (Balanced Half-Samples),
sections 5.1 to 5.9

- Method and theory

- Examples of application
- Partial balancing/combined strata

Class: 6
Day: Feb 12, 2013
Lecturer: Fay
Topic: Chapter 5. Balanced Repeated Replication (Balanced Half-Samples),
section 5.10 to 5.15

- Application when there are more than two PSUs per stratum
- Fay’s adaptation
- Software (WesVar, R)

5
Class: 7
Day: Feb 19, 2013
Lecturer: West
Topic: Chapter 6. The Jackknife

- Method and theory

- Examples of application
- Combined strata and other methods of reducing the number of
replicates
- Computer software for balanced repeated replication and the
jackknife (WesVar, R)

Chapter 7. The Bootstrap

- Method and theory

- Examples of application
- Bootstrap confidence intervals

Class: 8
Day: Feb 26, 2013
Lecturer: West
Topic: Chapter 8. Application of Replication Methods in Practice

- Replicate survey weights

- Replication of weighting adjustments (nonresponse adjustments,
poststratification, raking, generalized regression estimators)

Chapter 9. Comparisons among Methods for Complex Estimators

- Bias and precision of variance estimation

- Confidence interval coverage
- Availability and accessibility of software
- Ease of incorporating effect of weighting adjustments
- Use by secondary data analysts

Chapter 10. Inferential Statistics from Complex Survey Data

- Tests of significance
- Model parameters
- Use of variance estimation methods in inferences via the pseudo-
likelihood approach
- Generalized design effects for multivariate statistics

6
Class: 9
Day: Mar 5, 2013 (Michigan break, but class will be held)
Lecturer: Fay
Topic: Chapter 11. Linear Regression, section 11.1 to 11.6

- Incorporating effects of complex design on inference

- Consequences of ignoring complex design
- Accounting for complex design in practice

Class: 10
Day: Mar 12, 2013
Lecturer: Fay
Topic: Chapter 11. Linear Regression, section 11.7 to 11.9

- Tests for significance for model fit

- Computer software (SUDAAN, WesVarPC, R)
- Significance levels and confidence intervals for model coefficients

Day: Mar 19, 2013

UMD Spring Break—no class

Class: 11
Day: Mar 26, 2013
Lecturer: West
Topic: Chapter 12. Logistic Regression

- Taylor Series linearization for implicit (i.e., iteratively obtained)

estimators
- Standard errors for coefficients
- Tests of alternative models
- Computer software (SUDAAN, WesVar, Stata, R)

Class: 12
Day: April 2, 2013
Lecturer: Fay
Topic: Chapter 13. Analysis of Cross-Classified Categorical Data

- Goodness-of-fit
- Tests of association
- Measures of association
- Computer software (SUDAAN, WesVar, Stata)

7
Class: 13
Day: Apr 9, 2013
Lecturer: West
Topic: Chapter 14. Variance Estimation for Systematic Samples and One-PSU-
per-Stratum Designs

- “Measurable” versus “non-measurable” sample designs

- Paired strata/paired selection method
- Successive differences method
- Grouped selection method
- Model-dependent methods

Chapter 15. Generalized Variances - Modeling Sampling Errors

- Reasons for modeling sampling errors to generalize variances

- The relative variance approach
- The design effects approach

Class: 14
Day: Apr 16, 2013
Lecturer: Fay
Topic: Chapter 16. Survival Analysis

- Cox proportional hazards models—discrete and continuous time

- Kaplan-Meier estimate
- Software (SUDAAN, Stata, R)

Chapter 17. Multi-Phase Designs

- Examples of Multi-Phase Designs

- Variance estimation for multi-phase designs
- Software for estimating variances for multi-phase designs
- Distribution of take-home final exam

Day: Apr 23, 2013

Topic: Final Exam due, 5pm