The key to process improvement is Statistical Thinking!!
Data Mining leads to successful decision support!!
Statistically Designed Experiments result in productivity improvement
Core business of InduStat Pro
Data mining (analysis and training); Data analysis; Statistical Consulting Service; Software Training and Support (STATISTICA, DESIGN-EXPERT); Statistically Designed Trials (DOE) / Experimental design & data analysis; Statistical Process Control (SPC); Quality improvement.
Prediction and classification; Data Visualisation & Graphics; Data segmentation; Statistical model building; Multiple linear & nonlinear regression; Logistic regression; MARS (multivariate adaptive regression splines); CART (Classification and regression trees); Neural networks; Support vector machine models; Fractional factorial designed experiments; Response surface methods; Central Composite Designs; Box-Behnken designs; Shewhart Control Charts & Cusum methods; multivariate control charts; Contingency table analyses;
close corporation (1996/039213/23, registered for VAT: registration
· Founded in 1996 by Dr Nico Laubscher as a vehicle through which to provide a professional statistical consulting service to industry and other information-age focussed companies or research oriented organisations. The key focus of InduStat Pro is on the solution of statistical problems recognising the importance of statistical knowledge in quality and productivity improvement. As InduStat Pro has proven expertise and a fundamental understanding of the concept of variability, it is exceptionally well placed for problem solving in a data rich environment.
· A focus as important as problem solving is that of training in the industrial aspects of statistics, viz. Data-mining, Design and analysis of experiments and Statistical Process Control. Successful workshops in data mining, SPC and DOE have been run and are continuously being revised and improved. Workshops in aspects of statistical science integrated with its application in Statistical Software (STATISTICA) have been provided for several large companies. Workshops will be tailor-made in cooperation with several customers. Typical clients were Department of Labour (Pretoria), ESKOM, Botswana National Productivity Centre, Hillside Aluminium, British American Tobacco, The University of Stellenbosch: Process Engineering, The University of the Western Cape: Department of Chemistry, and others.
(inter alia) in the following aspects of data analysis and general statistical
building via linear or nonlinear regression, the analysis of variance, and
other specialised areas such as correspondence analysis, forecasting & time
series. Another of InduStat's major activities is the design of cost
effective trials for process improvement (two-level and mixed-level fractional
factorial designs, Central composite designs, Box-Behnken designs, etc.). Statistical process control
(Shewhart and CUSUM charts, multivariate control charts, etc.) is a statistical
tool in which InduStat Pro has innovative knowledge.
Uses the following software for data analysis and presentation of
results: Microsoft Office2007 (Professional), STATISTICA
8 Data Miner, StatXact 5
for Windows and Design-Expert 7.1. Program development work is done in Visual Basic for Applications or Statistica Visual Basic.
Pro is the sole authorised representative in Southern Africa for Stat-Ease
the producers of Design-Expert software, the leader in the field of the design and analysis of experimental
Every organisation that believes firmly in the principle of continuous improvement of systems and processes within systems, should establish links with a professional statistical consultant so that he or she can become an integrated part of all relevant teams working toward product and process improvement. Such a relationship should aim at a long-term commitment from both sides. The statistician should have nothing else in mind but adding value to the products and/or processes through his or her contribution.
Statistical consultants should supply unique technical support on aspects of planning improvement designs, ensure that plans are implemented correctly and in line with recognised statistical principles of designed trials, perform required analyses of the data obtained, document the statistical part of the result and ensure that documentation contains clear-cut interpretation and recommendations so that management has explicit and well-formulated information on which to base corporate decisions.
The statistician should become well-versed in the operations and culture of the companies served through continuous participation in all relevant activities and should strive towards becoming recognised and well-respected as part of the success story of the company, earned through hard work, timeous delivery of results and sympathetic perseverance at times when the path to follow is not clear at all.
If the vision set out above, is a general one that many statisticians may like to claim for the statistical profession in general, our mission is a singular one: to do everything possible - with professional efficiency - to ensure that the vision becomes a reality in the InduStat Pro consultancy. We would like to be recognised for the effort put into obtaining solutions to the problems arising from the mandates of our customers. The solution of problems will be pursued ceaselessly and without digressing from the remit and time-frame laid down. The interests of each individual customer, in particular the safety and integrity of their data with which we are entrusted, will be protected at all costs.
Membership of professional societies:
o Fellow of the South African Statistical Association (ex-President and ex-editor of the S.A. Statistical Journal): 1957 – present.
o Member of the American Statistical Association: 1974 – present.
I started my professional career as a statistician at the South African CSIR in December 1956, directly after completing an M.Sc. degree in mathematics at North-West University, Potchefstroom (known as Potchefstroom University at that time). In 1960 the degree of Doctor Scientiae was awarded to me by my alma mater on the thesis "On transformations for the stabilization of variance and the normalization of distribution functions". I have never deviated from a career path in statistics and data analysis over a period of 52 years. Deming's concept of statistical thinking (one of the fundamentals of quality management), has been a major guiding light in the development of the skill of successful real world problem-solving that I now consider to be my forté. In addition, successful scientific, business and engineering problem-solving for the statistical scientist can never be achieved without teamwork. I consider part of the success that I have achieved, as due to being able to co-operate and communicate successfully with teams of scientists, engineers and business people.
Following a career of 19 years with the CSIR, I filled positions as
professor of Statistics at the Nelson Mandela Metropolitan University (known at
the time as the University of Port Elizabeth) in the years 1976 – 1986, and the
University of Stellenbosch (1987 - 1989). For the seven years 1989 - 1996, I
worked as Company Statistician in the manufacturing environment of SANS
Fibres (Pty) Ltd,
Thus I have had ample exposure to both the university research/teaching environment and the hard world of problem solving in an industrial environment. In 1996 I founded and since then successfully managed InduStat Pro as a statistical consultancy. I maintain highly professional ethical standards and proprietary information is safe with InduStat Pro. Total secrecy of proprietary information is guaranteed and insight obtained through work for one customer shall never be used to the unfair benefit of another.
Statistical expertise stretches across the wide spectrum of statistical analyses, including applications to the fields of engineering, business, biological and medical statistics. A strong foundation in theoretical statistics is considered to be a necessary requirement for successful solution of difficult practical statistical problems. Such expertise exists at InduStat Pro.
Technical publications, reports and presentations
1. (With Kidd, M.), (1993), "Robust estimation of scale based on Hampel's lemma five with application to the Rayleigh family", S. A. Statist. Jour., Vol. 27, pp. 203 - 220.
2. (1994), "Comments on the use of spectral control charts", Technometrics, Vol. 36, pp. 335 - 336.
3. (1994), "How to approximate a histogram by a normal density", The American Statistician, Vol. 47, pp. 251-252.
4. (With Kidd, M.), (1995), "Robust confidence intervals for scale and its application to the Rayleigh distribution", S. A. Statist. Jour., Vol. 29, pp. 199 - 217.
5. (1996), "A variance components model for statistical process control", S. A. Statist. Jour., Vol. 30, pp. 27 - 47.
6. (With Kidd, M.), (1998), “A review of decision tree methodology to determine the structure of large data sets”, IMT Technical Report: Simon’s Town.
7. (With Kidd, M.), (1999), "Bayesian Belief Networks and its application”, IMT Technical Report: Simon’s Town.
8. (With Kidd, M.), (2000), "Methodology and Application of Multivariate Adaptive Regression Splines (MARS)”, IMT Technical Report: Simon’s Town.
9. (2001), “Neural Networks for Pattern Recognition”, Powerpoint presentation for De Beers Group: Cape Town.
10. (With Kidd, M.), (2002), "Association analysis and Genetic Algorithms: An introduction to two rule based induction methods”, IMT Technical Report: Simon’s Town.
11. (2002), “Rule Induction by Association Analysis in Data Mining”, Conference in honour of professor Francois E. Steffens”, University of South Africa, Pretoria. (pp. 85 – 98).
12. (With Kidd, M.), (2005), One-day workshop in Data-mining, South African Statistical Annual Conference, Rhodes University: Grahamstown.
13. (2006), Designed experiments for improving productivity. Invited presentation: Department of Physics, University of the Western Cape.
14. (2008), Text mining. A presentation in the Seminar series at the Statistics Department of the University of Stellenbosch.
15. (2009), Aspects of industrial statistics. Invitational presentation to the Stellenbosch Statistical Symposium (arranged by International Society for Business and Industrial Statistics. 26 August 2009.
1. Data Mining and Knowledge Discovery in Databases
This is a major continuing project in which the body of knowledge known as Data Mining (DM), is applied to large databases to explore and model relationships between variables. The idea is to turn huge amounts of data, possibly disorganised (perhaps originally designed with another purpose in mind), into useful information. The process moves through various phases starting with the formulation of the objective of the work, cleaning and transforming the data, sampling to establish the learning (training), validation and test data sets. The models to be analysed are selected and the tools to do the analysis, executing the analysis (the DM step) and to validate the findings are selected.
far as we know the first data mining workshop presented in South Africa was
been organised by the S A Statistical Association at its Annual
Conference in Durban (November 1999). The presenters were Nico Laubscher and
Typical tools used in the DM step include model building techniques like
multiple linear and non-linear regression, logistic regression,
discriminant analysis, cluster analysis, decision tree analysis, neural networks, Support Vector Machines and other modern tools
such as Boosting and
Details of the Data Mining training program may be downloaded as a pdf-file from the file download area.
2. Industrial Statistics Workshops
Another ongoing activity at InduStat Pro is the presentation of industrial statistics workshops. The general workshop covering the main branches of industrial statistics is aimed at giving a basic outline of useful methods in statistical process control (SPC) and design of experiments (DOE) in the spirit of what has become known as Six-sigma methodology. The idea is to transfer sufficient knowledge to the customer to become self-sufficient in the basic application of these methods.
In the beginning of 1997 InduStat Pro acted as a consultant for the Statistics Department at Rhodes University to help them in the design of courses integrating topics from quality management and industrial statistics into their under- and postgraduate curricula.
A major seminar was presented under the auspices of the S A
Statistical Association (at the Annual Conference, November 1997),
covering both Design of Experiments and Statistical Process Control. An honours
course in designed experiments was presented on behalf of the Department
of Statistics at the
A successful two-day seminar on statistically designed experiments and
response surface methodology has been presented (June 1998) to scientists,
engineers and sugar technologists of Tongaat-Hulett Limited and
of the Sugar Milling Research Institute,
Details of general information anout the DOE training as well as the training program may be downloaded as pdf-files from the file download area.
InduStat Pro is the appointed reseller in
One of the outstanding features (among many) of DX7 is the effective way in which stepwise regression models (forward, backward or regular stepwise) can be applied to main and interaction effects to build parsimonious optimal response surface models. Another is the intuitive way in which the Derringer & Suich algorithm, JQT 1980, for finding the sweet spot of a multiresponse process is implemented. In the case where only a few factors are involved, DX7 can overlay contour maps of the various responses in a single display, enabling the engineer to find the optimum operating conditions of the process intuitively, using subject knowledge. A recent innovation implemented in DX7.1 is the “Fraction of Design Space” (FDS) plot. This curve shows what fraction of the design space has a given prediction error or lower. This is a great tool for comparing various designs aimed at optimization.
Cost: For a cost estimate for DX7, please send an e-mail to industat.proBYmweb.co.za (replace BY with @). Specify the configuration you need, e.g. single corporate license, academic research license (only available to universities) or corporate network license (several seats for simultaneous use required).
The following is a list of some of the customers of InduStat Pro: The e-mail addresses cannot be used as direct links (just to make it somewhat more difficult for spammers). Please replace “BY” in any e-mail address with the customary “@”. Some of the individuals referenced below may no longer be in the positions or companies indicated.
Tongaat-Hulett (Pty) Ltd,
SANS Fibres (Pty) Ltd, Bellville (Mr D A Cowan: 021-959-4307; cowandBYsans.co.za )
Truworths (Pty) Ltd,
Department of Polymer Science,
Department of Industrial Engineering,
Provincial Administration: Western Cape. Department of Governmental Affairs & Housing (Mr R J Ellis: 021-483-4790)
S A Droëvrugteraad,
De Beers Mineral Resource Management, Wells (UK), (Mr C F Prins: +447834114683: chris.prinsBYdebeersgroup.com )
Old Mutual, Pinelands,
Arvin Exhaust SA (Pty) Ltd,
Crusader Systems (Pty) Ltd., Technopark, Stellenbosch (Dr J Ludik: 021-880-1677: jludikBYcrusader.co.za )
De Beers Consolidated Mines,
ESKOM Peaking Generation, Bellville (Mr R B Kydd: 021-914-3111: robert.kyddBYeskom.co.za )
Botswana National Productivity Centre, Gaborone, Botswana (Dr S. Taneka: )
Plant Protection Research Institute, Stellenbosch (Mr J A Gordon: 021-8874690: )
SAPPI Technology Centre, Springs (Ms M. Allaway: 011-360-0279)
Ground Hornbill Research & Conservation Project, Warmbad, (Mnr H Coetzee: 014-734-1788:
The information (registration forms and workshop programs) concerning training work-shops for data mining and for statistically designed experiments may be downloaded from the links on the right.
How to contact InduStat Pro
South Africa: 021-880-0147
This web page was designed by Nico Laubscher and updated on August 31, 2009.