Professor Goldstein is a chartered statistician, has been editor of the Royal Statistical Society's Journal, Series A, 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.
There are three main foci of my research interests.
I have started a new BLOG commenting on current educational issues. This is at HGeduc.blogspot.com . If you follow me on Twitter @GoldsteinHarvey then you will be told when a new commentary is posted.
The Department for Education has, for 2012 and 2013 produced more extensive information in the annual school league tables. Unfortunately, despite concerns being raised with them, there are some serious problems with these which can easily lead to misinterpretations. Click here for a note on this.
A report produced for the British Academy provides a comprehensive discussion of league tables in the public sector
.Recent work with George Leckie has demonstrated how the uncertainty associated with using school league tables for choosing schools, can be illustrated graphically in ways that make it very accessible to users. Click here. In this connection, some recent work by Allen and Burgess appears to provide evidence that league tables can make useful predictions, but on close examination this evidence turns out to be weak and based upon a questionable analysis. Click here for a critique.
A2010 paper in the British Educational research journal seeks to resurrect the Rasch (item response) model for educational testing, using rather dubious arguments. A detailed response to this, which the journal has refused to publish without even going through a refereeing process, can be viewed here.
New research on the uses of school league tables shows that for purposes of school choice they convey very little useful information.
One development of the REALCOM project has been to develop a full multilevel multiple imputation procedure that will handle normal as well as categorical data. This procedure has been implemented using an interface between REALCOM and MLwiN. The software and a manual can be downloaded. The methodology is described in a paper on mixed multivariate responses that has been published in Statistical Modelling. The software has now been updated to handle weighted imputations. The latest version will now write out chains for parameters selected for monitoring. This may be useful for studying functions of parameters. See the new manual for details.
Note: some of the documents on this page are in PDF format. In order to view a PDF you will need Adobe Acrobat Reader