DEPARTMENT of MATHEMATICS   Dr Stephan Rudolfer Honorary Research Fellow  in Mathematical Statistics Department of Mathematics University of Manchester

Tel:  +44 (0)161 275 5912 (direct line and voicemail) Tel:  +44 (0)161 275 5800 (general office) Fax: +44 (0)161 275 5819 (departmental) Email:  stephan@maths.man.ac.uk

Research Keywords
Carpal Tunnel Syndrome, Continuation Ratio Models, Discriminant Analysis, Electromyography, Logistic Regression, Medical Diagnosis, Performance Evaluation, Proportional Odds Models, Stereotype Models

Research Interests
Application of Discrete Discriminant Analysis to the medical field of Electrodiagnosis, especially the diagnosis of Carpal Tunnel Syndrome (CTS), using two important datasets:

 
1.  Huddersfield data: With the help and encouragement of the late Dr. John L. James, Consultant Physician, St. Luke's Hospital, Huddersfield, I have collected a rich and complex  medical dataset of about 1,000 patients involving the diagnosis of CTS.  It is fully described in Technical Report 2001/07, and is available from me as a text (ASCII) file of approximately 450 kb.

 
2. Canterbury data: Dr Jeremy D. P. Bland, FRCP, Consultant Neurophysiologist, Kent & Canterbury Hospital, has a steadily increasing database of over 12,000 patients, to which he has given me access since 2000.

• Comparison of various ordinal logistic regression models (Proportional Odds, Continuation Ratio, and Stereotype), with investigation of the statistical problems thrown up by this model fitting.

• Investigation of the problem of variable selection (around 100 covariates are recorded to facilitate diagnosis of CTS, only a few of which are likely to have a significant influence on the diagnosis), initially with the binary diagnosis (CTS / non-CTS).

• Performance evaluation of the fitted models using, for example, the following criteria:
• Brier's Score
• Cohen's Kappa Statistics
• Concordance Index, c
• Goodman & Kruskal's Gamma Statistic
• Percentage Correct
• Somers' D Rank Correlation Indices
• Model validation
• Bootstrap
• Cross-validation
• Data Splitting
• Resubstitution
• Comparison of approaches to non-responses to electrical stimulation as part of the diagnosis of CTS
• Appropriate discretisation of the continuous responses, with non-response as an extra category
• Conditional probability arguments
• Bivariate models
• CTS is a bilateral condition, whereby the probability of CTS in one hand is increased if the other hand already has CTS.
• This means that the CTS diagnoses of a patient's hands are not independent.
• Accordingly, the most appropriate models are those fitting the bivariate response variable (Left Diagnosis, Right Diagnosis).
• Much work still needs to be done to apply such models to CTS diagnosis.
• Sequential models
• In fitting statistical models, the usual method is to use a design set to obtain the parameters' estimates.  However, with an ever-increasing dataset, as patients present themselves to the clinic, it would save considerable time and effort if the existing estimates could be "updated" by a new cohort of patients, rather than refitting the whole, augmented, dataset.  Such sequential methods are well-known in Multivariate Normal Analysis, but do not exist in the logistic regression context.

Collaborative Work
Since 2000, I have been working with Dr Jeremy D. P. Bland, FRCP, Consultant Neurophysiologist, Kent & Canterbury Hospital,  on the medico-statistical problems associated with the diagnosis of CTS.  In particular, we have carried out an in-depth investigation of the age distributions of CTS patients in the two datasets, which turn out to be surprisingly similar: see Technical Report 2003/01.

Technical Reports (Manchester Centre for Statistical science)

2002/01:  Diagnosis of Carpal Tunnel Syndrome using Logistic Regression(2) (Lectures given at the Facultad de Matemáticas, Universidad Autónoma de Yucatán, Mérida, Yucatán, Mexico, 18-22 February, 2002).

2001/07: PATIENTS.DAT, a dataset involving the diagnosis of Carpal Tunnel Syndrome.

2001/01: Diagnosis of Carpal Tunnel Syndrome using Logistic Regression (Lectures given at the Facultad de Matemáticas, Universidad Autónoma de Yucatán, Mérida, Yucatán, Mexico, 12-15 February, 2001).
 

Publications

(with Georgios Paliouras and Ian S. Peers) A comparison of logistic regression to decision tree induction in the diagnosis of Carpal Tunnel Syndrome.  Computers and Biomedical Research 32 (1999), 391-414.

(with P.C. Watson and E. Lesaffre) Are ordinal models useful for classification? A revised analysis. J. Statist. Comput. Simul. 52 (1995), 105-132.

(with P.C. Watson) Evaluation of orthant probabilities for singular bi- and trivariate Normal distributions. J. Statist. Comput. Simul. (1993) 48, 219-232.

(with P. C. Watson) Table Erratum: Formula 26.3.16, Abramowitz, M. and Stegun, I. A.  Handbook of mathematical functions with formulas, graphs and mathematical tables, Wiley (1964), Mathematics of Computation, 59 (1992), 727.

CTSS: An interactive microcomputer program for the clinical screening of Carpal Tunnel Syndrome.  II. Statistical and computational aspects. Electromyogr. Clin. Neurophysiol. 30 (1992), 483-489.

CTSS: An interactive microcomputer program for the clinical screening of Carpal Tunnel Syndrome.  I. Clinical aspects.
Electromyogr. Clin. Neurophysiol. 28 (1988), 259-262.

Membership of the following Societies:

• Association of Christians in Higher Education at Manchester (Treasurer 2003-2006)

• London Mathematical Society

• Royal Statistical Society

• Chairman of the Manchester Local Group  since its founding in 1995
• Member of the Local Organising Committee (2001-2004) of the 2004 International Conference
• Joint organiser, with Prof. Graham Dunn, of the annual joint meetings of the RSS Manchester Local Group and the BiostatisticsGroup, Manchester University: October 2000, October 2001, October2002, October 2003, October 2004

Journal Referee

• Biometrics
• Journal of Statistical Computation and Simulation

Other Information

Chernobyl Children's Lifeline

Note: the pictures linked to this section of my webpage were taken  by Victor Mizzi of Chernobyl Children's Lifeline, and are reproduced with his permission.

In 2002, I did a sponsored walk with St Paul's Anglican church, Heaton Moor, Stockport, for this marvellous charity, receiving a total of Ł500.

In 2003, I did the same again, receiving a total of Ł950.

Many of my colleagues contributed to both these amounts, for which many thanks!

Unfortunately, the suffering continues, with even the second generation having similar terrible problems.
 

Hands-Only Stations of the Cross