Dr Sally Stares

This course is available on the MPhil/PhD in Computational Social Science, MPhil/PhD in Economic Geography, MPhil/PhD in Environmental Economics, MPhil/PhD in International Relations and MPhil/PhD in Regional and Urban Planning Studies. This course is freely available as an outside option to students on other programmes where regulations permit. It does not require permission.

This course is available to all Research students. This course is not controlled access. If you register for a place and meet the prerequisites, if any, you are likely to be given a place.

Basic descriptive and inferential statistics to the level of MY451 or equivalent. Familiarity with research design in the social sciences to the level of MY400. MY455 is not required but is strongly recommended.

This course provides an introduction to the design and analysis of population surveys. It is intended both for students who plan to design and collect their own surveys and for those who need to understand and use data from existing large-scale surveys. Topics covered include concepts of target populations, survey estimation and inference; sampling error and nonsampling error; sample design and sampling theory; methods of data collection; online surveys; responsive sample designs; interviewing; cognitive processes in answering survey questions; longitudinal and comparative surveys; design and evaluation of survey questions; nonresponse error; survey weights; and analysis of data from complex surveys. The course considers the current and future impact of Generative AI on survey workflows throughout, and students will gain practical experience of AI deployment.

The course is structured around two forms of survey inference. The first is from questions to constructs: we infer from observed phenomena (respondents' answers) to unobserved phenomena (the construct being measured). The second is from sample statistics to population parameters. Both forms of inference are unified within the Total Survey Error framework introduced in Week 1.

Software: Qualtrics for questionnaire design; R or Stata for survey estimation. No prior knowledge of Qualtrics or R is assumed.

20 hours of lectures and 10 hours of seminars in the Winter Term.
2 hours of lectures in the Spring Term.

This course has a reading week in Week 6 of Winter Term.

Five seminars run across the course. Seminars 1 and 2 form a linked pair on questionnaire design and evaluation. Seminar 3 is a sample design practical. Seminar 4 is a hands-on data analysis session in R. Seminar 5 is a methodological mystery in which students diagnose sources of error in a real set of discrepant survey estimates using the TSE framework. Formative coursework tasks are associated with Seminars 1, 3, and 4, and provide essential preparation for the summative assessment and examination.

Groves, R M, Fowler, F J, Couper, M P, Lepkowski, J M, Singer, E, and
Tourangeau, R (2009). Survey Methodology (2nd ed.). Wiley.
Biemer, P. et al (2017) Total Survey Error in Practice. Wiley.
Tourangeau, R, Rips, L J, and Rasinski, K (2000). The Psychology of Survey Response. Cambridge University Press

Exam (70%), duration: 120 Minutes in the Spring exam period.

Project (30%).

• Take-home assessment (30%): Covers question design, evaluation, and estimation. Deadline: 3pm Thursday of Week 11 (exact date confirmed at start of term). Students work independently on a structured problem set applying concepts from across the course.
• In-person examination (70%): Covers all course material. Format and rubric provided in the first lecture.