Diploma/MSc in Official Statistics Professional Development Programme STAT 6097 Compensating for Non-Response Outline Programme Building 39, Room 2015, Southampton Objectives To provide students with an understanding of the sources and nature of nonresponse in surveys and its potential effects on estimation. To provide students with an understanding of alternative weighting methods used to compensate for the effects of unit non-response. To provide students with an understanding of alternative methods to compensate for the effects of item non-response, with a focus on imputation methods. To enable students to design a strategy involving weighting and/or imputation for dealing with non-response in a particular survey. Prior Background Assumed STAT6093 Survey Sampling is a pre-requisite and it will be assumed that students understand the basic principles of estimating finite population parameters from sample data. Some methods to be discussed will involve the application of linear and logistic regression, so some familiarity with regression to the level of STAT6095 Regression Modelling will be assumed. Learning Outcomes Students will achieve an understanding of: the sources, nature and effects of non-response the main approaches to weighting for unit non-response the main approaches to imputation for item non-response how to choose between alternative weighting or imputation methods for a particular survey Syllabus Types of non-response Response rates Reasons for non-response Prevention of non-response Weighting for unequal selection probabilities Sample-based weighting
Population-based weighting Calibration weighting methods Bias and variance properties of weighting methods Weight truncation and standardisation Designing a weighting strategy Non-respondent sub-sampling and substitution Approaches to item non-response Imputation methods Bias and variance properties of imputation Variance estimation in the presence of imputation Designing an imputation strategy Modelling approaches to missing data Introductory Reading in Context of UK Government Elliot D (1999) Report of the Task Force on Weighting and Estimation. GSS Methodology Series No 16. Elliot D (1997) Software to weight and gross survey data. GSS Methodology Series No 1. Report of the Task Force on Imputation (1997). GSS Methodology Series No 3. Each of these three reports may be downloaded from http://www.ons.gov.uk/ons/guide-method/method-quality/specific/gss-methodologyseries/index.html Introductory Reading on Methods Lehtonen R and Pahkinen EJ (1996) Practical Methods for Design and Analysis of Complex Surveys. Second Edition, Chichester: Wiley, Chapter 4. Lohr S L (2010) Sampling: Design and Analysis. 2nd Ed. Brookes-Cole, Chapter 8. References for Course Bethlehem J, Cobben F and Schouten B (2011) Handbook of Nonresponse in Household Surveys. Hoboken: Wiley. [wide-ranging overview. Chapter 8 on weighting and Chapter 14 on item nonresponse are particularly relevant]. Cox BG et al (1995) Business Survey Methods. New York: Wiley. [Chapters 22 and 25 provide good descriptions of imputation and weighting methods, respectively, for business surveys.] Groves R, Dillman DA, Eltinge JL and Little RJA (2002) Survey Non-Response. New York: Wiley. [contains introductory chapter on weighting by Bethlehem (ch.18) and some useful chapters on new developments in imputation, especially chapters 20-22.] Lessler JT and Kalsbeek WD (1992) Non-Sampling Error in Surveys. New York: Wiley. [Chapter 8 contains fairly compressed accounts of weighting (pages 192-199) and imputation (pages 210-229)]
Little RJA and Rubin DB (2002) Statistical Analysis with Missing Data, 2nd Ed. New York: Wiley. [Chapter 3 provides a brief account of weighting. Chapter 4 contains slightly more on imputation.] Särndal, C-E and Lundström, S (2005) Estimation in Surveys with Nonresponse. Chichester: Wiley. [Developed from a Current Best Methods manual at Statistics Sweden. It has good coverage of both weighting and imputation, with an emphasis on putting into common calibration framework. This is probably the most useful single text.] Blackboard Site All students are automatically enrolled on STAT6097-28308-14-15: 14-15- Compensating for Non-Response-28308 blackboard site. If you do not have access to the site, please let your tutor know so that she can enrol you. There is also a discussion board on Blackboard so that you can communicate with your fellow students. To submit an electronic version of the coursework assignment for the 'Turnitin' plagiarism software, you must access the specific module STAT6097-28308-14-15: 14-15-Compensating for Non-Response-28308 on Blackboard. Once you locate the specific module, you need to upload the assignment through the 'Assignments' folder according to the instructions listed below. Only students submitting coursework assignments will gain access to this site. If you don t have access to the site, please let your tutor know. NOTE: the uploading of an electronic version of the coursework assignment does NOT replace the two hard-copies of the assignment that are to be sent to the MOffStat administrator using the coursework hand-in form available at: http://www.southampton.ac.uk/ demography/postgraduate/taught_courses/msc_official_statistics.page? Instructions for uploading assignments: To submit coursework electronically via the TurnitinUK plagiarism device, login to the Blackboard unit for this module and select the Assignments link from the left-hand menu. Find the 'assignment' and click View/Complete. There will be a series of screens to complete, and full instructions. Please note the following points: 1. During the submission process you will see a receipt screen confirming that your assignment has been submitted. This will show the title of your paper and will also provide a Paper ID. You should make a note of this Paper ID for future reference in case of any problems. 2. Following your submission through Blackboard you will receive an email confirmation that your coursework has been received. This usually takes about an hour but may take up to 24 hours depending on how busy the system is. You should retain a copy of this email which will act as a receipt for your electronic submission. If you do not receive this email then you should contact the module coordinator as soon as possible.
Assessment The assessment will be by coursework involving two questions. The first question will require you to perform various weighting schemes to estimate the population mean for a specified sample. The second question will be about imputation. The deadline for submission of the coursework is Tuesday, 6th January, 2015 with the postmark required by the preceding Monday. Please send 2 hard copies of the assignment to the MOffStat Administration. The address is: Claire Woodley University of Southampton Building 58, Room 2101 Highfield, Southampton SO17 1JB Students who fail to meet the deadline will receive a 5% reduction in their mark for every working day up to 5 working days. After 6 working days, the standard policy is to award zero for the coursework assignment. Procedure for coursework extensions If students want to request an extension, they must have medical or personal circumstances to justify the late submission of assessed coursework (medical evidence must be substantiated by a doctor's note). They should initially fill out a coursework extension request form available at the above website and send it to their Personal Tutor, Dr Solange Correa-Onel, who will, where appropriate, agree to the extension of the deadline after discussion with the module co-ordinator and examination officer. Under no circumstances will extensions be allowed beyond a period of two weeks. Resit arrangements No mark below 25% is allowed for any module in the programme. All students receiving below the minimum mark will have to resit the module according to School of Social Sciences policy. Academic integrity The University places the highest importance on the maintenance of academic integrity in the conduct of its affairs, and the Academic Integrity Statement for Students can be found in the University Calendar available online at http://www.calendar.soton.ac.uk/ sectioniv/academic-integrity-statement.html. Please familiarise yourself with what is expected of you in this regard by reading through this statement. Your attention is drawn particularly to Appendix 1, which outlines those things which you must seek to avoid, including cheating and plagiarism. A very useful set of interactive guides is available at http://www.studyskills.soton.ac.uk. These aim to help you gain a better understanding of academic integrity and develop your skills so that your assessed work does not accidentally plagiarise the work of others.
You must take particular care in using sources in essays and in your dissertation. Remember that plagiarism includes not only verbatim copying but also direct paraphrasing of a source. Verbatim quotes from a source should always be in quotation marks, with the source indicated, and should be used only occasionally in an essay or other report. Detailed advice on appropriate referencing in essays and dissertations is given in the Department of Social Statistics Guidelines on writing essays and there is an online guide to referencing at http://library.soton.ac.uk/socialstats/referencing. Students are encouraged to discuss and exchange ideas, since this is an important part of the educational process. However, it is not acceptable that you read and gain ideas for your coursework from another student s finished work. Copying includes using another student's computer program, output or graphics. If academic integrity is deemed to have been breached, there are a range of penalties that may be applied. If you are unsure about what is and is not permitted, ask - we will be happy to explain and discuss. Troubleshooting Please get in touch with your personal tutor, Dr Solange Correa-Onel, should any difficulties arise with this module.
Outline Programme Monday 24 November 2014 10.00-10.10 Welcome and introduction to course 10.10-11.10 Introduction to non-response 11.10-11.30 Break 11.30-13.00 Introduction to weighting 14.00-15.20 Sample-based weighting Exercise 1 15.20-15.40 Break 15.40-17.00 Population-based weighting Exercise 2 Tuesday 25 November 09.30 10.50 Using models to construct weights Calibration methods 1 11.10-13.00 Calibration methods 2 Exercise 3 14.00-15.30 Using weights 15.30-15.50 Break 15.50-16.50 Exercise 4 (computer-based) Wednesday 26 November 09.30 10.50 Dealing with item non-response Imputation methods Exercise 5 11.10-13.00 Review of exercise Properties of imputation methods 14.00-15.20 Multiple imputation 15.20-15.40 Break 15.40-17.00 Edit and imputation Thursday 27 November 09.30-10.50 Statistical modelling methods I 11.10-13.00 Statistical modelling methods II 14.00-15.30 Exercise 6 (computer workshop on imputation) 15.30-15.50 Break 15.50-17.00 Private study Friday 28 November 09.30-10.50 Some developments in non-response 11.10-12.30 Discussion of coursework, Q&A session, review and evaluation of course