Around seventy friends and colleagues attended a memorial for Jon Rasbash on Friday 10th September 2010. Eleven speakers shared their personal memories of Jon which together built up a strong impression of Jon's enormous character and his devotion to people and to his work.
An early career prize and PhD bursary were announced at the memorial in honour of Jon's incredible contributions to quantitative social science, and his commitment to supporting young researchers. The national prize of £500 will be awarded biennially for achievement in the development and/or application of advanced quantitative methods in any social science discipline. There will be a call for nominations in 2011. The annual bursary of £2000 will be granted to a Bristol postgraduate student undertaking research in advanced quantitative methods.
CMM have released an informal film of Jon Rasbash, not originally intended for publication, speaking about the history of MLwiN at the MLwiN 20th anniversary party, University of Bristol, March 2009. Short film
The Centre for Multilevel Modelling is seeking to appoint two social statisticians or social scientists with advanced quantitative skills for two ESRC-funded research projects:
We would like to announce our latest MLwiN release - 2.20. Some minor new features/bug fixes have been introduced, eg, Fixed bug in MOVE command with non-default column names, also MCMC seed menu option now applies to macros too. For further details of the features and bug fixes go to bug fixes
Realcom and Realcom-Impute applications have been updated. REALCOM = "Developing multilevel models for REAListically COMplex social science data". Realcom and Realcom-Impute are free downloadable applications to enhance MLwiN under certain conditions - for further details go to our Realcom and Realcom-Impute pages
Some bugs have been fixed, eg, a problem with loading text columns from Stata data files. More about bug fixes…
We have also upgraded our free 30 day trial version to MLwiN 2.19.
In the UK, School League Tables are used extensively by policymakers and parents as one of the key measures of school effectiveness. However it has been widely recognised that there are problems with some of these measures. The first day of this two day event includes talks from a range of leading international experts and will cover a range of methodological issues around the measurement of value added, school performance and use of school league tables. Further details…
The Centre for Multilevel Modelling is very pleased to announce the addition of Stata practicals to our free on-line multilevel modelling course. These give detailed instructions of how to carry out a range of analyses in Stata, starting from multiple regression and progressing through to multilevel modelling of continuous and binary data.
A selection of sample PDF documents which can be viewed without the need to log in:
The Stata practical tutorials are complete and ready to view, though some of the quizzes refer only to MLwiN. We are currently working on revised versions of the quiz questions to make them more relevant to Stata users. We will let you know when we have completed these.
Example of a Stata practical (you will need to register/log in):
Stata P 3.1: Regression with a Single Continuous Explanatory Variable:
This course is free. We ask only that you complete a registration form if you have not done so already, as fully as possible, to help us to conduct research into learning multilevel modelling and to improve our materials. We hope that this is a valuable resource for anyone who would like to learn more about multilevel modelling and how to apply these methods using MLwiN and Stata.
Good luck and do let us know if you have any problems registering, logging on or with any aspect of the learning materials.
MLwiN 2.18 is now available. Some minor new features/bug fixes have been introduced, e.g. Allowed copy/paste of categories between columns, and Fixed potential problem when adding/removing categories from a column. Go to bug fixes
Upgrade to latest version:
It is with regret that we have to announce that the CMM director, Jon Rasbash has passed away. We would like to offer our condolences to his family and to his many friends and colleagues. Jon will be sadly missed.
Jon Rasbash, who was Professor of Computational Statistics and Director of the Centre for Multilevel Modelling at the University of Bristol, sadly passed away on 10 March, 2010, aged 49. He is survived by his wife, Jacky and children, Zoe and Molly.
Jon was an outstanding researcher with a truly international reputation and at the peak of his career, He was also a wonderful person, a wise and generous manager, and a friend to all those who had the good fortune to share in his work. The loss to social and medical science is immense. Jon was involved with many things, but principally he had become known for his development of multilevel methodology and its software implementation and for his interest in studying social relationships within families.
Jon was born the youngest of four children to David and Maureen Rasbash in November 1960. He spent his later childhood in Edinburgh where his father was appointed as Professor of Fire Engineering at the University. He then studied at the University of York and left university in 1982 with a degree in Biology and Computer Science. He then went to work for Professor Bill Hill in the Department of Animal Genetics in Edinburgh and published his first paper with Professor Hill on modelling long-term artificial selection in finite populations in 1986. His first introduction to medical research was his work with Professor Michael Healy at the London School of Hygiene in 1986-1987 with whom he published two papers in the growth studies in Annals of Human Biology. Following that he moved around the corner to the Institute of Education to work with Professor Harvey Goldstein on multilevel modelling, a methodology then in its infancy. The group, joined by Bob Prosser, Huiqi Pan, Min Yang, Ian Plewis and Geoff Woodhouse, began to develop not just the methodology and training materials but, crucially, under Jon's supervision, the software, later to become MLwiN, that was required to fit the statistical models. Joined later by Bill Browne and Fiona Steele the Centre for Multilevel modelling was formed and eventually found its present home in Bristol. There were numerous developments during this period, in all of which Jon played a key role, and without his support, enthusiasm and insight, many things would not have happened.
The move to Bristol saw Jon take over the direction of the Centre and in the comparatively short time of 5 years, with the support of all his colleagues in the centre team, at Bristol, he succeeded in establishing it as one of the University's elite groupings with large amounts of research income, numerous publications and considerable amounts of material to support users of multilevel models. He was very active in organising workshops for a wide range of users, all over the world. His enthusiasm and knowledge inspired many.
His most ambitious project, well under way at the time of his death, was the development of a general framework for the specification of complex statistical models that would enable users, with different levels of statistical and computing expertise, to set up complex models that would interface with a variety of software packages. This work, and much of the other activity he was involved in will be continued by the Centre team and his other collaborators, all of whom are grateful for the support and inspiration that he has given.
Jon was totally unique. Those of us who had the privilege of knowing and working with him have been vastly enriched and whatever research emerges over the coming years will have Jon's imprint all over it. Thanks for everything Jon. We will miss you.
Harvey Goldstein, Fiona Steele and William Browne.
Bristol, March 2010.
New versions of Realcom, Realcom-Impute and Matlab Installer for Realcom are now available.Note: If you downloaded the Matlab Installer before 02-Mar-10 you will need to update your installation.
We would like to announce our latest MLwiN release - 2.17. Some minor new features have been added, e.g. Allow copying results from the "variance" window, and also some bugs have been fixed. For details of the features and bug fixes go to bug fixes.
We are pleased that our MLwiN Forum is becoming increasingly valuable and some excellent answers to questions about MLwiN and multilevel modelling are appearing on there. Even if you think your question may not be a good one, do log on and post it anyway, we cannot guarantee anything but it may be worth a try: www.cmm.bristol.ac.uk/forum.
We recently launched a Train the Trainers initiative to increase UK capacity for delivering training courses in multilevel modelling. If you are a UK academic and would like to give your own course in multilevel modelling, we might be able to offer some support in developing your materials.
Have a Happy New Year from all of us at the Centre for Multilevel Modelling!
If you are a UK academic and would like to give your own course in multilevel
modelling, we can provide some support in developing your materials.
Applications must be received by 26 February 2010.
This is also to let you know that there are only 3 places left on our
'Introduction to Multilevel Modelling in MLwiN' course, April 2010. Further
Some bugs have been fixed, e.g. preserving whether a term is in the fixed part when loading old worksheets.
Our free MLwiN 30-day trial version has now been updated to MLwiN 2.16.
2) The Centre for Multilevel Modelling's George Leckie and Harvey Goldstein
have had a paper published in the Journal of the Royal Statistical Society:
Series A, 172(4), 835-851, '172, Part 4, pp. 835-851, 'The limitations of using
school league tables to inform school choice'.
Each year, the government publishes league tables of GCSE results to help
parents choose their children?s secondary school. But as George Leckie and
Harvey Goldstein explain, the past performance of schools is an imprecise
guide to how they might perform in the future.
Speaking about the findings, George Leckie said:
"Parents need to be aware that the tables contain less information than
official sources imply and that this necessitates a lower weight being
placed on them as compared with other sources of information available to
parents. It is also worth noting that Scotland, Wales and Northern Ireland
have either never had or have abandoned publishing school league tables.
Now seems a good time for England to follow suit."
For further details go to the Journal of the Royal Statistical Society:
Series A, 172(4), 835-851
3) ESRC +3 PhD Studentship in Social Statistics. Application now closed
New versions of the REALCOM and REALCOM-IMPUTE applications are now available with fixes to known bugs. Further details…
At a recent conference in Cambridge organised by Longview - a think tank set up in 2005 to promote longitudinal and life course research and development (www.longviewuk.com) - it was decided to establish a new international learned society - the 'Society for Longitudinal and Life Course Studies'. The unique feature of the Society will be its multi-disciplinary nature, bringing together members across the health sciences, behavioural sciences, social sciences and methodology, with a shared interest in longitudinal and life course research. In addition to the opportunity to bridge disciplinary boundaries, the Society will offer to its members:
An Interim Executive Committee has been set up to build membership and organise elections for the Honorary Officers of the Society and the Executive Committee. The Committee is moving ahead with membership recruitment straight away. You can become a Foundation Member of the society for which applications must be received by the end of January 2010.
We are offering a special privilege to Foundation Members:
In the interim period of establishing the Society, Longview, which is a Registered Charity and Company Limited by Guarantee, will receive all membership fees to be held in Trust on behalf of the Society. Once the Executive Committee is established and the Society is in a position to manage accounts the funds will be transferred. Fees can be paid in a convenient exchange currency, without paying a premium for transfer, i.e. in Euros, £s or $s.
Go to www.longviewuk.com.
The Society's membership year runs from beginning of February to end of January. For the first year we plan to invoice all registered applicants for their fees by the end of January 2010. But if you prefer to pay your fee now please do so on-line, or by BACS or by cheque, using the form supplied at the above web address. A direct debit facility (from your bank account) will be available in due course.
Once membership is established at a level of at least 50 paid-up members, the Interim Committee will be inviting nominations for the Honorary Officers of the Society and membership of the Executive Committee. The election process itself will take place (on-line) in the month of February through to the end of March. The new Committee will be formally ratified at a first business meeting in April 2010.
All members of the Interim Executive Committee see this as an exciting venture, and world-wide support for the Society will ensure its success. We therefore look forward very much to receiving your application.
Very best wishes,
Michael Wadsworth (Longview), John Bynner (Longview), Harvey Goldstein (U of Bristol), Paul Boyle (U of St Andrews), Amanda Sacker (U of Essex), Ingrid Schoon (Institute of Education, London), Dimitri Mortelmans (U of Antwerp), David Blane (Imperial College, London) , Carli Lessof (National Centre for Social Research, London) , Elizabeth Webb (Imperial College, London), Matthias Richter (U of Bielefeld)
1) We have now released MLwiN 2.15.
Some bugs have been fixed, e.g. we have fixed a potential error when reading
worksheets containing column descriptions. Further details about MLwiN 2.1 bug
fixes go to our bugs page
2) We are delighted to announce the long-awaited arrival of the Centre for
Multilevel Modelling's own MLwiN Forum. We would like to invite you to
participate in this forum, asking or answering questions about the software.
There are only a few questions there so far which we have put up as tasters, so
please go ahead and use it as the more activity there is the greater the
knowledge base for everyone.
3) With apologies - our January workshop 'Introduction to Multilevel Modelling
in MLwiN' has had to be moved to April 2010. We have now rescheduled this one
to be the week before our 'Multilevel Modelling of Discrete Response Data'
workshop, with just a weekend between them.
'Introduction to Multilevel Modelling in MLwiN' is now 7-9 April 2010
'Multilevel Modelling of Discrete Response Data' - 12-14 April 2010 - now open
This latest version of MLwiN has been updated to work with the new weighting feature of the latest REALCOM-IMPUTE. Also fixed reading in Stata files after there was already data in the worksheet.
A new version of REALCOM-IMPUTE is now available. In addition to providing a very general multilevel multiple imputation procedure for mixtures of normal and categorical variables it now allows the use of weights, such as those used in sample surveys, within the imputation. An updated manual is also available.
Topics covered include
* Multilevel data structures and alternative analysis strategies
* Introduction to MLwiN & revision of multiple regression
* Variance Components Models
* Random Intercept Models
* Random Slope Models
* Contextual effects
* Three-level Models
* Modelling variance
As a result of your feedback we have improved the clarity of the grading system - you can check your grades here (you will need to log into the course to see this page):
We are always interested in feedback about the web site and/or our on-line course. Please let us know of any possible improvements, errors or any other feedback - email me at email@example.com
We are pleased to announce our latest version of MLwiN: version 2.13. There are two important features in the latest release ( see item (1) below…):
(2) Other news: New RSS Special Issue devoted entirely to the theme of multilevel modelling. ( further details …)
(3) Fiona Steele, Professor of Social Statistics at the Centre for Multilevel Modelling - the youngest ever female Fellow to be elected to the British Academy: see item (3)
You can now upgrade for free if you already own a previous version of MLwiN (or are eligible for download - see conditions here: )
The Journal of the Royal Statistical Society has published a new Special Issue devoted entirely to the theme of multilevel modelling.
The editorial is completely free to download as a PDF file: Recent advances in multilevel modelling methodology and applications
Professor Steele's research in multilevel modelling, longitudinal data analysis and demography has led her to collaborate with social scientists and economists on projects covering areas including health and finance in developing countries and the impact of school resources on pupil attainment in England and Wales. She said: "My election is a great honour. As a statistician, it means a great deal to me that my contribution to social science has been recognised."
She will become the academy's youngest ever female Fellow and the youngest overall for more than 60 years. Established by Royal Charter in 1902, the British Academy champions and supports the humanities and social sciences. It aims to inspire, recognise and support excellence and high achievement across Britain and internationally.
We are pleased to announce the addition of a new module to our online training course: Module 7 on Multilevel Models for Binary Responses. This module has two integrated components: a description of concepts and models (‘concepts’) and detailed instructions of how to carry out analyses in MLwiN and interpret the results (‘practical’). Further details on the estimation of multilevel models for binary responses are given in a technical appendix. There are also interactive quizzes to test your understanding.
A new online journal devoted to longitudinal studies has just published its first issue. The International Journal of Longitudinal and Life Course Studies carries substantive and methodological articles and is free for both readers and authors. It has a distinguished panel of international referees and seeks to promote high standards of research in this area. To visit and register go to www.journal.longviewuk.com/index.php/llcs .
Throughout 2008, Prof. Herb Marsh, Ms. Alison O’Mara, and Dr. Lars-Erik Malmberg conducted meta-analysis workshops and seminars, funded through the ESRC Researcher Development Initiative. These materials are freely available for you to download. Please acknowledge the relevant authors if you use them.
We are pleased to announce the addition of a new module to our online training course: Module 6 on Regression Models for Binary Responses. This module has two integrated components: a description of concepts and models (concepts) and detailed instructions of how to carry out analyses in MLwiN and interpret the results (practical). There are also interactive quizzes to test your understanding.
Module 7 on Multilevel Models for Binary Responses will be available soon.
We have fixed some bugs on MLwiN 2.10, you can now upgrade to to our latest version - MLwiN 2.11. Our latest releases begin with 2.1… and are the same as 2.10 except that bugs have now been fixed. If your version is not the latest one you can upgrade in the usual way, and you can also check which bugs have been fixed. The training materials work with all MLwiN versions from 2.10 onwards.
We are pleased to make available a new free piece of software MLPowSim that is designed for performing sample size/power calculations in multilevel models via simulation.
The software package has been developed as part of a UK ESRC funded-project and is an 'old-fashioned' text input program that creates files that can be used in conjunction with MLwiN or R to perform the necessary computations to perform complex power calculations.
The program is available along with an extensive 150 page manual from my web page
We will describe the software currently as a Beta version as we have only had time to do preliminary testing and we haven't included much error trapping. The software is FREE and as such comes with no guarantees in terms of producing correct answers (although we hope it does!) and no guarantee of fast response to fixing of any bugs reported (although we hope there aren't many). We will however be genuinely pleased if people use it and let us know of any bugs they find or if they have a 'wish list' of additional features they might like.
Good luck with the software,
We had a gathering of colleagues and friends to celebrate MLwiN's (and MLwiN's precursors, ML2, ML3 and MLn) 20th birthday on Wednesday 4th March 2009.
Many of CMM's collaborators attended the birthday party and enjoyed the MLwiN cake:
The MLwiN 20th birthday cake
During the celebrations, Jon Rasbash spoke briefly about the beginning of the software, and related that it was first created during the early days of his collaboration with Harvey Goldstein (who is still part of the CMM team) and Michael Healy. Micheal Healy is now in his 80s, and as a young man worked with the famous statistician, Sir Ronald Aylmer Fisher, this lineage bringing to mind the saying 'standing on the shoulders of giants'.
The CMM team moved from the Institute of Education in London to the Graduate School of Education at the University of Bristol in 2005, a move which has been a great success and has cemented our collaboration: "It's been great for us to relocate here, we are all really happy, the Graduate School has been a great home for us as has BIPA (the Bristol Institute of Public Affairs) and the wider university. We are working closely with colleagues from other disciplines such as Geography, Economics, Veterinary Science and Social Medicine".
Thank you, Becky, for coming up with the birthday celebration idea, and Michelle for organising it so well. It is important for the team to stop once in a while and review where we are and how far we have come. This is especially important as usually the team focuses on looking to the future, developing exciting and pioneering methods in multilevel modelling.
Jon finished by saying the Graduate School of Education will be "a great place to spend the next 20 years"
Jon Rasbash cutting the MLwiN cake with Chris Charlton who now does much of MLwiN programming (on the left) and Tim Bond, Head of the Graduate School of Education Department (right)
Research Workshop in Multilevel Modelling, 7th-9th Sep 2009
This is a new workshop, adapted from our popular 'Getting Started' workshop. It provides an opportunity for a small group of researchers to analyse their own multilevel datasets with support from the CMM team, and to get advice on how to interpret and present their findings. For further details on this workshop and how to apply for it, go to our workshop page.
The Centre for Multilevel Modelling team is pleased to announce the release of MLwiN 2.10. This is the full non-beta version. You can upgrade to MLwiN 2.10 for free if you own an existing 2.02 licence.
If you are a UK-academic and eligible for a free download (more info about eligibility here) you can upgrade in the usual way*.
Updated web and print versions of the manuals are available here
To purchase the software or bound copies of the manuals, go to:
You can learn about multilevel modelling and how to use MLwiN on our on-line course:
There are step-by-step modules on using MLwiN, for example, 'Comparing Groups using Multilevel Modelling':
www.cmm.bris.ac.uk/lemma/mod/lesson/view.php?id=276&pageid=337 (you will need to log into the course to view this page).
*as long as you have 2.02 on your machine, if you only have one of the 2.10 beta versions, you may need to reapply for the full version as the upgrader works on the basis of you already having 2.02.
Tragically, Ken Rowe, a good friend and colleague, died defending his home in Marysville, near Melbourne, in the recent wildfires.
Ken was a talented researcher who made important contributions in a number of areas, notably multilevel structural equation modelling, and in the analysis of educational test data. Over many years he devoted himself to promoting good research practice, and multilevel modelling in particular, to audiences in Australia and across the world.
He had considerable charm and his enthusiasm was contagious. Thank you Ken for everything you did. Along with your family, we shall miss you enormously.
We have now released MLwiN 2.10 Beta (10), the last version the before (non-beta) MLwiN 2.10 which will be ready very soon.
You can upgrade to MLwiN Beta 2.10 (10) for free if you have an existing licence or are eligible for a free download. If you have already downloaded your MLwiN software (free-qualifiers or purchasers) you can also upgrade for free.
Don't forget you can upgrade to the non-beta version when it is released but there is a technical requirement that you should be aware of:
Note to all users:
You must have MLwiN 2.02 installed on your machine for upgrades to work automatically.
For further details about this and other questions you may have about MLwiN, please see our recently-expanded FAQ page.
We have found a bug in MLwiN 2.10 Beta versions 1-9 which may be very important to some users.
We have found in some rare cases that MLwiN 2.10 Beta (subversions 1-9) produced incorrect likelihoods for models estimated with (R)IGLS. The problem does not apply to models estimated in MCMC (since there is no likelihood calculated for these models; the problem does not apply to the DIC). The problem was not present in previous release versions (2.02 and earlier). The problem will be fixed in MLwiN Beta 2.10 release 10, which will be available shortly.
Our research has found the problem to only occur in single level models with a large number of cases (>10,000). Furthermore, the problem does not occur if a second level is declared but no variance component is fitted at level 2. If users have fitted a single level model in MLwiN 2.10 Beta (subversions 1-9), with only a single level hierarchy defined, the -2*log(likelihood) should be regarded as suspect and the model should be re-run in MLwiN 2.10 Beta subversion 10 when this becomes available.
We apologise for any inconvenience this may have caused our users. If you have any queries about this please contact our support team on firstname.lastname@example.org.
1) 2009 CMM Workshops
New 2009 workshops are now open for bookings. For further details, go to our workshop page
*** Introduction to Multilevel Modelling, 25-27 March 2009 ***
This workshop provides an introduction to multilevel modelling. We assume that participants are familiar with single level regression models (to the level implied by Module 3 of our online multilevel modelling course) but have no prior knowledge of multilevel modelling. Topics covered include
* Multilevel data structures and alternative analysis strategies
* Introduction to MLwiN & revision of multiple regression
* Variance Components Models
* Random Intercept Models
* Random Slope Models
* Contextual effects
* Three-level Models
* Modelling variance
Theory sessions are accompanied by practical sessions using our software, MLwiN, in which participants get the chance to apply what they have learned to real datasets. Again, no prior knowledge of MLwiN is assumed: the first practical takes participants right from the very basics. Throughout, there is an emphasis on how to interpret the models and on what kinds of research question they can be used to explore.
Prices: £85 for student
£145 for academic
£550 for non-academic
*** Research Workshop in Multilevel Modelling, 7-9 Sep 2009 ***
There are clear pre-requisites and a limited number of places for this
specialist workshop, and we will have to consider each application
individually. Places will only be offered to participants who can
1. a clear idea of the research questions they wish to explore with multilevel modelling,
2. a suitable dataset which they have already analysed using standard 'single-level' techniques, and
3. some experience in the application of multilevel models.
Prices: academic: £400, non-academic: £600
*** Two RSS workshops: Multilevel Modelling, 23-24 April 2009, and also 26-27 November 2009 London ***
For statisticians and quantitative researchers with an interest in the analysis of hierarchically structured data. Previous participants have included university researchers from the social sciences, medicine and public health and government statisticians. More details about the RSS workshops can be found at the RSS web site.
2) MLwiN - dollar and euro price reduction
Now may be a good opportunity to make the most of reduced GBP exchange rates as we have reduced our dollar and Euro prices for MLwiN. Also VAT is now down to 15% and this applies to some buyers - see our ordering page:
3) Our on-line learning materials are proving very popular. We are pleased to announce a new addition to the materials: a piece by Prof
Harvey Goldstein on 'The use of performance indicators in education'. It is now on-line here (you will need to log in to access the learning materials):
We are always pleased to hear any feedback about our learning materials and the web site. Please email the admin team on email@example.com with any comments or enquiries.
1) *New MLwiN 2.10 Beta (8) available*
You can now upgrade to MLwiN Beta 2.10 (8) (Vista-compatible) for free if
you own an existing MLwiN software licence, or are a UK academic and are
eligible for a free download: /cmm/software/mlwin/download/upgrades.html
You can also purchase the new MLwiN Beta (8) software if you do not qualify
for the free download. Subsequent MLwiN 2.10 Beta versions can then be upgraded by you directly for free once you have purchased or downloaded.
2) *Realcom-Impute now available*
Professor Harvey Goldstein is pleased to announce a new Realcom (=Developing
multilevel models for REAListically COMplex social science data)
application available - Realcom-Impute.
Realcom-Impute will, in conjunction with MLwiN, carry out a full multiple
imputation procedure. There is a full description in the supplementary
manual for MLwiN version 2.10 - /cmm/software/mlwin/download/manuals.html
3) Three new RSS Multilevel Modelling workshops to take place in London are now available for booking. Please email firstname.lastname@example.org for further details of the RSS workshops.
4) We have announced one new workshop in Bristol which is not yet open for bookings to take place in September 2009: 'Research Workshop in Multilevel Modelling'. However you can email email@example.com to request a place on the waiting list.
5) PLASC/NPD User Group Workshop: 18/19 November 2008, Bristol -
This two day international meeting is the final one in the current
PLASC/NPD User group Workshop series.
Further details of all these workshops can be found at:
6) The presentations from our recent 'Introduction to Multilevel Modelling' workshop can now be downloaded from our web site: /cmm/software/support/workshops/#introsep
*New MLwiN 2.10 Beta (7) available*
You can now upgrade to MLwiN Beta 2.10 (7) (Vista-compatible) for free if you own an existing MLwiN software licence, or are a UK academic and are eligible for a free download:
Question: Am I eligible? See: UK-academic
If you have already downloaded your MLwiN software (free-qualifiers or purchasers) you can upgrade for free here.
You can also purchase the new MLwiN Beta (7) software if you do not qualify for the download. Subsequent MLwiN 2.10 Beta versions can then be upgraded by you directly for free once you have purchased or downloaded.
*Two new video presentations within the LEMMA learning materials*
Please note, you will need to register on our learning materials site to view these videos unless you have already done so, just the click the links above and you will be taken through the simple registration process.
Bookings are now open for our Multilevel Modelling of Discrete Response Data workshop to take place in January 2009.
academic staff: £145**
**This price applies to all nationalities and includes a set of manuals (one user and one MCMC) lunch and refreshments every day, plus one evening meal at a local restaurant.
This intermediate level workshop provides an introduction to multilevel modelling of discrete (categorical and count) data. We assume that participants are familiar with multilevel models for continuous response variables (to the level implied by our Introduction to Multilevel Modelling workshop or Module 5 of our online multilevel modelling course) and standard (single-level) models for binary responses. Topics covered include:
* Multilevel models for binary response data
* Estimation procedures and software (with an emphasis on MCMC estimation in MLwiN)
* Models for proportions
* Multilevel multinomial logit regression for nominal responses
* Multilevel proportional odds models for ordinal responses
* Multilevel Poisson regression for count data
For further details go to our workshop page.
*School league tables: what can they really tell us?*
New research findings by Professor Harvey Goldstein and George Leckie, based in the Centre for Multilevel Modelling, were broadcast on this Monday's edition of BBC Radio 4's Learning Curve.
Harvey Goldstein spoke to presenter Libby Purves about how recent research, taking the GCSE marks of students in England and comparing the 'value added' scores with simple GCSE averages, has concluded league tables were not fit for that purpose. Harvey highlighted it was time that their publication should cease:
The research has demonstrated that for purposes of school choice, a government priority, these tables convey almost no useful information since they are essentially concerned with predicted future performance and that turns out to be extremely imprecise. The study used the government's National Pupil Database to carry out the analyses.
The results of the study are especially relevant at the moment since a number of voices have been raised in opposition to league table publication and targets on the grounds that these introduce great stress, distort learning and undermine good teaching.
Harvey and George hope their work will influence educational policy and practice and provide sound statistical evidence to influence the government, which has often said that its policies are driven by evidence. This will be a good test of that claim.
To listen to the interview, download the full research findings, visit:
*Covariance and Correlation Matrices*
We have just published another in our popular series of slide presentations with voice-overs: Covariance and Correlation Matrices.
We use multilevel modelling when we have dependent data, i.e. there is similarity between observations from the same group (for example, heights of children from the same family). An obvious question is: just how does the multilevel model take this dependency into account? In this presentation, we examine the structure of the model: we see what the correlation is between each pair of level 1 units in our dataset. This allows us to see how the relation between different observations from the same group is specified by the model. We contrast this to a single level model, for which we see there is no correlation between different observations from the same group.
See our video presentations page for more details about this and other audio-visual presentations from the Centre for Multilevel Modelling.
If you are interested in audio-visual learning materials you might like to know that there are two new slide presentations which have been synchronised with voice-overs and subtitles on our web site:
Do go to our video presentations page for details about these and other audio-visual presentations from the Centre for Multilevel Modelling.
We are very pleased to announce that MLwiN 2.10 Beta is now available!
This latest version of MLwiN will import and export Stata, SPSS, and Minitab data files. It also has an improved model specification interface, a new bespoke predictions window, and surface plotting routines. See more details about the new and improved features.
If you already own a licence for MLwiN 2.02 or have downloaded a copy as a
UK-academic, you can now upgrade to the new Beta version for free:
MLwiN 2.10 Beta will automatically be installed in a different folder from the old version, so you can continue to run both versions if required.
Our other big news is that our on-line multilevel modelling course is now available on the Centre for Multilevel Modelling web site: 'LEMMA' (= Learning Environment for Multilevel Methodology and Applications). The course contains a set of graduated modules starting from an introduction to quantitative research progressing through to multilevel modelling of continuous data.
From here you can log in and begin using the extensive learning materials
for free. We would be grateful if you could find the time to complete
the brief personal profile on your first log-in which will help us
in our research, and to help our funders evaluate the course. See our
You can choose any module or progress from Module 1 right through to
Module 5. We currently have five modules available but plan to add to
these in time:
1. Using quantitative data in research
2. Introduction to quantitative data analysis
3. Multiple regression
4. Multilevel structures and classifications
5. Introduction to multilevel modelling
There are also quiz questions, an introductory video (we will put up more
videos over the next few months) and other useful resources.
You will need MLwiN 2.10 to do the exercises within the learning materials.
You can either upgrade your current version or you can download a limited
teaching version for free. The teaching version is available to all
whether you own a previous copy of MLwiN or not.
We have been working on the materials for some time, and very much hope you
will find them useful. Do let us know what you think of the course. You
can send us your thoughts via a feedback form on the learning site.
(You will need to log in to get to this as it is within the learning site
We started 2008 with an excellent Introduction to Multilevel Modelling workshop. We are sorry that many of you were disappointed that you didn't get on this or the Getting Started course in March as they have been very popular and now fully-subscribed. However, in the meantime we are currently developing extensive on-line training materials and these should be available on our web site in the next few months. We will let you know when these are launched on this newsletter, probably late February or March.
The Royal Statistical Society are running a 2-Day Training Course on Multilevel Modelling in London, presented by Harvey Goldstein and Fiona Steele on 19-20 May 2008, at The Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX (Nearest Tubes: Barbican, Liverpool Street, Moorgate, Old Street)
This course is designed to give participants a solid grounding in the theory and application of multilevel models. The course will be based around theoretical sessions followed by 'hands on' practical sessions illustrating the theoretical concepts. The hands on sessions will use the MLwiN software package.
Statisticians and quantitative researchers with an interest in the analysis of hierarchically structured data. Previous participants have included university researchers from the social sciences, medicine and public health, and government statisticians (e.g.DfES, Home Office, ONS). The course will not assume a high level of statistical knowledge, but participants should be familiar with the application and interpretation of multiple regression analysis.
Non-Fellows: £764.00 plus vat
Linked Associate Member: £654.00 plus vat
Fellow: £599.00 plus vat
CStats: £544.00 plus vat
Student Fellows: £324.00 plus vat
Contact: Ali Houghton, Email: firstname.lastname@example.org
1) We have recently published some new materials on the Centre for Multilevel Modelling web site, on the LEMMA page (Learning environment for multilevel methodology and applications):
(i) Harvey Goldstein has recently published a paper: Modelling the effect of Pupil Mobility on School differences in educational achievement.
Value added analyses of school performance typically assign students to the school where they take the outcome test or examination. A particular problem with this is that this ignores the fact that many pupils move school between the time that they take the prior test or examination and the time that they take the outcome test. For example, for GCSE outcome in England where an 11-year old test score is used to adjust for prior achievement, up to 40% may have changed school in some areas. Ignoring this will tend to reduce the amount of variation estimated at the school level and distort inferences in terms of school effectiveness. The paper demonstrates this effect, using multiple membership models with data from English Local Authorities provided by the National Pupil database. It also shows that rank orderings of schools are almost unaffected when the proper analysis is carried out.
(ii) What are the barriers to using more complex quantitative methods? and Fitting a two-level model- both presentations by Sally Thomas
(iii) The Multilevel Latent Covariate Model: A New, More Reliable Approach to Group-Level Effects in Contextual Studies by Oliver Lüdtke, Herb Marsh and others
(iv) Significance Testing
Powerpoint presentation with audio commentary and MLwiN demo by Kelvyn Jones
Tests for coefficients of individual variables: eyeballing standard errors, Wald tests, and calculating p-values using the tail areas screen in MLwiN. Tests for comparing models: the Likelihood Ratio Test and the Deviance Information Criterion.
More details on all of above
Further learning materials will appear on the web site soon.
2) Workshops: We are sorry that some of you have been disappointed as the Introduction to Multilevel Modelling course in January 2008 is already fully-subscribed. We cannot add any more of you to the waiting list as there are already 5 people on it. However, there are 11 places left on the Introduction to Multilevel Models: Getting Started with your own Data workshop in March 2008.
RECENT ADVANCES IN MULTILEVEL MODELLING METHODOLOGY AND APPLICATIONS
There will be a Joint Meeting of the Social Statistics Section/ General Applications Section
December 5th 2007 9.00 am to 5.30 pm at the Society, 12 Errol Street
London, EC1Y 8LX.
methods. The speakers will be
William Browne, James Carpenter, Mark Gittoes, Harvey Goldstein, George
Leckie, Jassy Molitor, Bengt Muthén, Ian Plewis and Anders Skrondal.
Topics include multilevel multiple imputation, efficient MCMC estimation of
discrete-time event history models, combining mismatched administrative and
survey data, diagnostics for generalized linear mixed models, the use of
multilevel modelling in higher education policy development, and ethnic
group differences in educational attainment. More details
With best wishes,
We are pleased to announce that we are now accepting bookings for our Introduction to Multilevel Modelling workshops to take place early 2008. We would advise early booking as these courses quickly become fully-subscribed.
Introduction to Multilevel Modelling: Tues 15 - Thur 17 January 2008
Designed to give participants a solid grounding in multilevel modelling. The workshop will be based around theoretical sessions followed by "hands on" practical sessions illustrating the theoretical concepts. The hands on sessions will be using the MLwiN software package. Prices:
academic (staff): £145
Introduction to Multilevel Models: Getting Started with your own data: 31 March - 4 April 2008
An introduction to multilevel models combining theory and hands on practical sessions with provided data sets. 7 hours of sessions are allocated for participants to put into practice the concepts and techniques they are learning on their own data set. Prices:
3) We are now running the web site for the Avon Local Group of the Royal Statistical Society, an an opportunity for statisticians in Bath and Bristol to meet and discuss statistical methods and their applications in areas of local interest, including healthcare, education, and the environment. Please contact Nicky Welton (Nicky.Welton@bristol.ac.uk) to reserve a place.
Information about our workshops and conferences is available here
We have three news items, free MLwiN 30-day trial software, Realcom bug-fixes and announcement of RSS conference. You can be updated with CMM news by RSS news fee.
1) We can now offer a free 30-day MLwiN download. The software is the full version, identical to the original product except that it will expire after 30 days. If you haven't already tried MLwiN then this might be a good chance for a brief evaluation.
2) Following Harvey Goldstein's three successful workshops presenting the new Realcom software, (REALCOM = Developing multilevel models for REAListically COMplex social science data), as is often the case with new software there have been some initial bugs. However Harvey has now developed a bug-free version and has updated the manual accordingly. You are welcome to test this out and do of course let us know if you find any further bugs.
3) Announcement of RSS day conference in London: Joint Meeting of the RSS Social Statistics and General Applications Sections: 'Recent Advances in Multilevel Modelling Methodology and Applications', 5 December 2007, 9am-5pm.
Retired/Student/EDA Fellows £25
Linked Associates £45
None of the above £60
(includes lunch and morning tea/coffee)
The presentations are planned as follows:
- The use of centered parameterisations and MCMC estimation to fit discrete time survival and event history models, William Browne, Martin Green, Mousa Golalizadeh and Fiona Steele
- Analysis of multilevel datasets with missing observations: links between multilevel models and multiple imputation, James Carpenter
- Multi-level modelling: an essential analysis?, Mark Gittoes
- Modelling multilevel multivariate response data for arbitrary distributions, Harvey Goldstein
- Bayesian graphical models for combining mismatched administrative and survey data: application to low birth weight and water disinfection by-products, Nuoo-Ting (Jassy) Molitor, Chris Jackson, Sylvia Richardson, Nicky Best
-Multilevel Mixture Modeling Methods, Bengt Muthén, Tihomir Asparouhov
-Ethnic group differences in educational attainments and progress revisited, Ian Plewis
-Diagnostics and prediction for generalized linear mixed models, Anders Skrondal and Sophia Rabe-Hesketh.
1) We are seeking to appoint a statistician or quantitative social scientist to join an interdisciplinary research team specialising in the analysis of social data with complex structure to address important substantive research questions. The researcher will be based in the ESRC-funded project Learning Environment for Multilevel Methodology and Applications (LEMMA) at the Centre for Multilevel Modelling. LEMMA is part of the National Centre for Research Methods (NCRM) which consists of a coordinating Hub at the University of Southampton and six Nodes spread across the UK which is charged with delivering a step change in social science research methodology. Closing date for applications: 20 July 07.
2) RSS Multilevel Modelling workshop in June: This course was designed to give participants a solid grounding in the theory and application of multilevel models. Presentation slides now downloadable from our workshop materials section.
Dear Multilevel Modellers
The Royal Statistical Society are running a Multilevel Modelling two-day course presented by Harvey Goldstein and Fiona Steele. This RSS PDC 2-Day course is designed to give participants a solid grounding in the theory and application of multilevel models. The course will be based around theoretical sessions followed by hands on practical sessions illustrating the theoretical concepts. The hands on session will use the MLwiN software package.
Further information and a registration form on this and other courses by the RSS PDC.
One place only in London on 15th June and two in Bristol on 3rd July are available on the 'Realcom: Recent advances in multilevel modelling' free workshop. Please book as soon as possible if you would like to attend as the places will probably soon be taken.
Thank you for your interest in the Centre for Multilevel Modelling. We have had many new requests from you asking to join our mailing list. Welcome to all, and do let us know if you details are incorrect or you no longer wish to receive emails (see contact details below)
Three news items: 1) advanced methods workshops (free) 2)date for 2.1 release and 3) web site statistics: details of our most popular pages
1) Three repeated 1-day workshops on advanced methods in multilevel modelling will be held at University of Birmingham, 12 June '07, Royal Society Of Tropical Medicine & Hygiene, 15 June '07, University of Bristol, 3 July '07. These workshops are FREE and include lunch and refreshments.
The Centre for Multilevel Modelling at the University of Bristol is completing a three year ESRC supported research project developing new methodology for fitting complex multilevel data. The research team currently consists of Harvey Goldstein, Jon Rasbash, Fiona Steele and Christopher Charlton. The project has developed methodology and associated training materials in the following areas of multilevel modelling: structural equation models, measurement errors and multivariate mixed response types (Normal, and categorical) at more than one level of the data hierarchy. A full description of the research grant application can be found on our Realcom page.
The methodology builds upon that already implemented in MLwiN version 2.02. The routines use MCMC estimation and are written as free-standing MATLAB modules (http://www.mathworks.co.uk/) with graphical user interfaces for setting up models and displaying results. There is a set of training materials that has been written that provides an introduction to the methodology and a guide to using the software. Applications are to a variety of problems, including flexible prediction models, multiple imputation for missing data in multilevel models, latent variable models for test scores, and misclassification errors in social status data.
The MATLAB software routines (or a free standing EXE for users who do not have a copy of MATLAB) and training materials will be available for download for those registered for the workshops.
2) The Beta version of MLwiN 2.10: currently scheduled for release around August 2007 but this date may be subject to change, we will keep you informed.
3) Web site statistics: We are very pleased that our new web site launched last autumn has proved increasingly popular. In the last seven months we have had over 39,000 unique visitors (based on cookie information) and more than 150,000 page loads. The number of visitors show a steady increase over this time. Professor Jon Rasbash has been delighted to see his video/PowerPoint presentation 'Why use Multilevel Modelling' (What is it and why you should do it) has been consistently in the top ten most frequently-visited pages since it was published:
(Internet Explorer-only at this point though more widely-compatible videos will be coming soon)
The other most popular pages after the home page and the MLwiN page have been relatively recent additions to the web site, all learning resources:
i) What are multilevel models and why should I use them?
ii) Types of Model
iii) Multilevel Structures and Classifications
We welcome any feedback, good or bad about the web site and look forward to providing more on-line learning materials in 2007/08.
1) Two places have recently become available on our popular Getting Started with your own data workshop, 26-30 March 2007:
This workshop is designed as an introduction to multilevel models combining theory and hands-on practical sessions with provided data sets. In addition 7 hours of sessions are allocated for participants to put into practice the concepts and techniques they are learning on their own data set. Five tutors will be available for the practical sessions where participants are analysing their own data.
2) We have now published a detailed guide to Getting data into MLwiN including step-by-step help when various problems are encountered:
3) Professor Harvey Goldstein has finished making sure all his publications (except books) from 2000-2007 are now on his section of our web site and can be downloaded.
1) Workshops: We are pleased to announce that two workshops with places available in September can now be booked on-line.
i) Introduction to Multilevel Modelling, 10-12 September '07
This course is designed to give participants a solid grounding in multilevel modelling. The workshop will be based around theoretical sessions followed by "hands on" practical sessions illustrating the theoretical conc