Statistics - the key to good science

There is a lot of bad science around, which is reflected in Minerals Engineering's current rejection rate of 67% (posting of 26 November).  A recent report by the Nuffield Council on Bioethics showed evidence of scientists increasingly “employing less rigorous research methods” in response to funding pressures. A 2009 survey found almost 2% of scientists admitting that they have fabricated results; 14% say that their colleagues have done so.

The last statement is particularly worrying, but of the 67% of papers rejected by Minerals Engineering a very high proportion is due to lack of scientific rigour in the design of experiments and the interpretation of results. There appears to be a significant lack of understanding among many workers of how to apply statistical approaches to research, but thankfully I have a number of very experienced and respected researchers who I can turn to for advice in this area.

One of them is Prof. Tim Napier-Munn, who has recently delivered the 135th of his statistics course which he presents around the world. He is passionate about our profession utilising statistics properly and has now had a book published, based on his professional development courses, which should be on the bookshelf of every researcher in our industry.

Written by a mineral engineers for mineral engineers, Statistical Methods for Mineral Engineers (How to Design Experiments and Analyse Data) is packed with real world examples, and de-mystifies the statistics that most of us learned at university and then forgot.  It shows how simple statistical methods, most of them available in Excel, can be used to make good decisions in the face of experimental uncertainty.  Written in accessible language, it explains how experimental uncertainty arises from the normal measurement errors and how statistics provides a powerful methodology to manage that uncertainty.  It assumes only that the readers are numerate, can use Excel, and want to do a better professional job.  It is aimed squarely at mineral engineers and allied professionals (such as chemists) on the mine site, in head office, in engineering and supply companies and in universities.

Topics include:

•    the presentation of data – charts, tables and PowerPoint.
•    uncertainty in data – precision, accuracy, the normal distribution, sources of error.
•    comparing quantities using hypothesis tests such as the t-test, F-test, chi-square test, ANOVA,  non-parametric tests.
•    modelling using regression analysis, including linear, non-linear and weighted regression.
•    designing and analysing efficient experiments and plant trials.
•    time series analysis, including variograms and time series models.
•    multivariate analysis (PCA, clustering, binary logistic regression, MANOVA).
•    performance monitoring and optimisation, including statistical process control and EVOP.
•    statistics for chemists and mineralogists, mass balancing, sampling (Gy theory).
•    Monte Carlo and bootstrap methods.
•    a selection scheme to choose the appropriate statistical tool for the job in hand.

 

The book is available from the JKMRC website.