Workshops on Multilevel Modeling in Education & Social Science Research

Prof. Harvey Goldstein

Science of Learning Strategic Focal Research Area

Speaker: Professor Harvey Goldstein

Date: 5-7 June 2017 (Monday to Wednesday)

Workshop 1: Introductory Workshop on Multilevel Modelling 

Workshop 2: Workshop on Analyzing Longitudinal Data Using Multilevel Modelling Methods 

Workshop 3: Efficient procedures for handling missing data in generalised linear models 

(Registration closed)

Many kinds of data in education and the human sciences have a hierarchical structure. For example, students are assigned to classrooms within schools, which are further nested within different geographical regions and school types. Multilevel data structures are also found in longitudinal studies where an individual’s responses over time are correlated with each other.

The Faculty of Education Science of Learning Strategic Focus Area has invited Professor Harvey Goldstein to conduct 3 Workshops (June 5-7) on Multilevel Modelling:

Workshop 1. Introductory Workshop on Multilevel Modelling (1.5 days, June 5 whole day & June 6 am)

Participants will be expected to have a basic understanding and experience of applying and interpreting ordinary multiple regression models. A set of training materials to enable participants to refresh their knowledge of this topic is available via a download from the Centre for Multilevel Modelling (CMM) website in Bristol (

The workshop will be hands on and participants will be expected to provide their own laptop computers running windows 7 or later operating system. It will be possible for two participants to agree to share 1 machine for the practical sessions. Prior to the workshop participants will be asked to download and install MLwiN from the CMM website. A 30 day fully functional trial version can be downloaded by participants. The software will also be available on the University of Hong Kong network. Participants should also download the two manuals from the web site: ‘A users guide to MLwiN’ and ‘MCMC estimation in MLwiN’. These PDF files can be used ‘on screen’ during the course or participants may wish to print the following chapters:
Users Guide: Chapters 1-14
MCMC manual: Chaps 1-4

By the end of the workshop participants should be able to conduct their own, unsupervised, multilevel analyses with continuous or discrete responses and to be able to understand applied papers that use these techniques.

Provisional programme.
Day 1—June 5
On the first day participants will be introduced to the topic of multilevel modelling using examples, with group teaching interspersed with practical ‘hands-on’ sessions where participants will be guided through data analyses on their computers. The day will concentrate on models with Normally distributed responses.
09.30 – 10.00 Registration (refreshments)
10.00 – 10.30 Introduction to multilevel modelling
10.30 – 11.15 Random intercept and random coefficient (slope) models
11.15 – 11.30 Break
11.30 – 12.45 Practical session
12.45 – 13.45 Lunch
13.45 – 15.15 Residual estimates, hypothesis testing, complex level 1 variance
15.15 – 15.45 Break
15.45 – 16.45 Practical
16.45 – 17.30 Resume and discussion

Day 2—June 6
On the second day participants will be introduced to multilevel models for discrete response data, MCMC estimation, models for non-hierarchical data, and recent work on missing data. As on day 1, lectures will be interspersed with practical ‘hands-on’ sessions.
09.30 – 10.00 Resumé; further opportunity to raise issues from day 1.
10.00 – 10.30 Multilevel models with discrete responses I: binary response models
10.30 – 11.15 Introduction to MCMC (Markov chain Monte Carlo) methods: estimation and diagnostics
11.15 – 11.30 Break
11.30 – 13.00 Resumé and general discussion

Workshop 2. Workshop on Analyzing Longitudinal Data Using Multilevel Modelling Methods (June 6 pm)

This is a half day workshop introducing techniques for handling repeated measures of longitudinal data. It will assume that participants will have a basic understanding of multilevel modelling. The basic approach will be illustrated with the analysis of human growth data. While it is not hands-on, there will be an opportunity for some discussion of participants own data.

Provisional programme.
2.00 – 2.45 Repeated measures data and polynomial growth curves
2.45 – 3.15 Example of repeated measures of heights of children fitted using MLwiN
3.15 – 3.30 Tea
3.30 – 4.15 Extensions to the basic model
4.15 – 5.00 Discussion, including participants presenting their own data

Workshop 3. Efficient procedures for handling missing data in generalised linear models (June 7 am)

This is a half day workshop that will assume participants have experience of analysing data with missing values. It is not hands-on but participants are encouraged to talk about their own datasets.

Provisional programme.
9.30 – 10.15 Background: different kinds of missingness and implications
10.15 – 10.45 A simple example to illustrate different procedures for dealing with missing data
10.45 – 11.00 Coffee
11.00 – 11.45 Multiple imputation
11.45 – 12.15 Recent developments
12.15 – 12.45 Discussion and participants’ examples

Places are limited and registration is necessary. Registration Deadline: 26 May, 2017. Email confirmation will be sent to all successful registrants.

About the speaker

Professor Goldstein is a chartered statistician and is currently joint editor of the Royal Statistical Society’s Journal, Series A. He has been a member of the Society’s Council and was awarded the Society’s Guy medal on silver in 1998. He was elected a member of the International Statistical Institute in 1987, and a fellow of the British Academy in 1996. He was awarded an honorary doctorate by the Open University in 2001.