Statistical analysis with missing data using multiple imputation

Date:

28/01/2015 - 29/01/2015

Organised by:

University of Manchester

Presenter:

Jonathan Bartlett

Level:

Advanced (specialised prior knowledge)

Contact:

mark.elliot@manchester.ac.uk

Map:

View in Google Maps (M13 9PL)

Venue:

Humanities Bridgeford Street
University of Manchester
M13 9PL

Description:

Outline

In this course we begin by discussing the issues and problems raised by missing data, and introduce the key concepts required for classifying missing data mechanisms into one of three types. We then consider some of the frequently adopted ‘ad-hoc’ approaches for handling missing data, and discuss their limitations. Next we introduce the method of multiple imputation, a practical and principled approach for handling missing data. Through computer practicals using Stata, participants will learn how to investigate missingness in their data and how to apply the statistical methods introduced in the course to realistic datasets, such as the National Child Development Study.

Objectives

Provide an introduction to the issues raised by missing data, and the associated statistical jargon (missing completely at random, missing at random, missing not at random)
Illustrate the shortcomings of ad-hoc methods (eg mean imputation) for handling missing data
Introduce the method of multiple imputation as a practical and principled approach for handling missing data.

Prerequisites

Participants should have a working knowledge of STATA, and in particular be familiar with regression models, such as linear and logistic regression.

The course is designed for researchers involved in social science and epidemiology who face the problem of missingness in their data analyses.