Methods for Missing Data: Multiple Imputation and Inverse Probability Weighting
Venue: Highfield Campus, University of Southampton, UK
Presenter: Dr James Carpenter
Dates of Course: Monday 3rd - Wednesday 5th November 2008
This course has already run. Please check the course listings for a future course.
Summary of Course:
The focus of the course will be on methods to adjust for item nonresponse in sample surveys, which occurs when some survey participants choose not to answer certain questions in the survey.
The course will have a strong practical focus with regular computer sessions to consolidate the ideas presented in the lectures, and to gain experience with the various methods.
(The course will not focus on unit-nonresponse which occurs when certain sample members do not take part at all in a survey. The CASS course ‘Handling Nonresponse in Sample Surveys', 1-3 June 2008, will address the problem of unit-nonresponse.)
The main objectives of the course are:
- To introduce the key concepts underpinning the analysis of partially observed data, together with a principled approach to the analysis;
- To explain the shortcomings of frequently used ad-hoc methods
- To introduce Mulitple Imputation, and gain familiarity with using the ICE software in Stata for multiple imputation, in simple and more complex examples
- To explore the role of sensitivity analysis, and methods for performing approximate sensitivity analysis
- To introduce inverse probability weighting and doubly robust estimation.
The course will consist of 6 sessions, each of which comprises a 1-h lecture followed by a discussion and then a 1.5h computer practical. The key topics covered will be:
- Session I: Introduction, issues raised by missing data, and towards a systematic approach
- Session II: Issues raised by missing data in experimental research, and shortcomings of ad-hoc methods
- Session III: Introduction to multiple imputation
- Session IV: Further issues in multiple imputation
- Session V: Sensitivity analysis
- Session VI: Inverse probability weighting and doubly robust estimation.
Computer workshops will enable course participants to put the learned methods into practice. The course will use the packages Stata (80%) and Mlwin (20%).
Further information about the contents of the course will be available shortly.
The course is aimed at researchers who face problems of missing values in some of the survey variables they wish to use in their data analysis. Participants may be academic researchers in the social and health sciences or may work in government, survey agencies, official statistics or the private sector. While the focus will be social, parallels and illustrations from medical applications will also be included.
Participants on this course must have prior statistical knowledge covering inference and linear regression modelling (up to the level of the CASS course on ‘Introduction to Linear Regression Modelling' for example).
The course will have a strong practical emphasis, with regular computing sessions, using STATA and MLwin, to enable participants to work through examples. Therefore, prior experience of analysing survey data using a statistical package is required. The practicals will contain all the details necessary for novice users; nevertheless basic familiarity with both packages is an advantage.
Participants will receive written course notes.
James Carpenter is a Reader in Medical and Social Statistics at the London School of Hygiene & Tropical Medicine (University of London). His main research interest is missing data, and developing and applying multiple imputation. He has led missing data courses in the UK and abroad, including under the ESRC's Researcher Development Initiative.
£25 per day for UK-registered students. £50 per day for staff from UK academic institutions (including research centres), ESRC funded researchers and registered charity organizations. £195 per day for all other participants. The course fee includes course materials, lunches and morning and afternoon tea, but not accommodation and travel which is to be arranged by the participant.
Location and Accommodation:
The course will be held at the Southampton Statistical Sciences Research Institute, Building 39, University of Southampton, Southampton, SO17 1BJ.
Participants are left to book their own accommodation according to individual needs. Further information on location and accommodation can be obtained from here.
The course will begin with coffee and registration at 9.30 a.m. on the first day and finish at 5.00 p.m. on the last day.
- Participants would benefit from reviewing the introductory material on www.missingdata.org.uk (under Getting Started)
- See also Kenward, M. and Carpenter, J (2007): Muliple Imputation: Current perspectives. Statistical Methods in Medical Research, 16, 199-218
Deadlines and Refunds:
Course places are limited and early registration is recommended. Payment must be made when submitting the registration form. Refunds for cancellation are as follows: Full refund for cancellation one calendar month before the course, no refunds can be made for cancellations after this date. Please note that in case of cancellation an administration charge of £30 may apply.