The ideas behind Stat-JR (pronounced "stature") were the brainchild of Jon Rasbash and have been developed since September 2009 by Bill Browne and his team as part of the ESRC nodes eStat and Lemma II. Stat-JR's objective is to produce tools to cater for different types of user: from the novice statistical practitioner to the statistical guru and algorithm developer. To do this, Stat-JR provides, under a unified interface, a modular system composed of templates designed to:
Stat-JR has been developed using the Python package with the view that statistical users of all abilities could contribute to the development of new templates to answer specific questions. As part of the software, we have developed an in-house algebra system which automatically produces the algebraic formulae for the posterior distributions of model parameters based on the model definition and user inputs. These algebraic formulae form the backbone of the MCMC algorithm. The system then converts these formulae into computer code (Python, C++) to be compiled before running the model. The code is always accessible, enabling the interested user to download and modify it within their own application.
This version offers the user access to a range of model templates (reproducing all the examples from the MLwiN MCMC manual, as well as other model types), data exploration templates and plotting templates. It also enables interoperability with a variety of other packages, including MLwiN, WinBUGS, JAGS, OpenBUGS, R, SPSS, Stata, AML, Minitab, Sabre, SAS, MATLAB, Octave, and Stan (via RStan), as well as an eBook interface. A few example eBooks are being distributed with the software, but additional eBooks and templates can be downloaded from here.
We successfully ran our first workshop entirely using Stat-JR in July in Manchester "Introducing multilevel models and applying them to the Health Survey for England using Stat-JR".
We will be running further workshops in the future and details will be posted on the CMM workshops page.
Stat-JR comes from a long tradition of writing software packages which started initially with the package NANOStat developed in 1981 by Professor Mike Healy at the London School of Hygiene and Tropical Medicine. This was the precursor of the MLwiN macro language. Jon Rasbash worked for Mike during this period and later transferred to the Institute of Education (London) working with Professor Harvey Goldstein. They were involved in development of software to handle multilevel data, namely:
Further developments included REALCOM to cover measurement error modelling, structural equation models, mixed responses and missing data imputations; and MLPowSim for power calculations in multilevel models.
The eStat project started in September 2009 and the first beta release of the Stat-JR software occurred in May 2012.
If you wish to contact us regarding: