Vision No. 9 Sep 2008

[Teaching version logo]

As a part of VSNi’s on-going commitment to supporting educators across the world, we have launched GenStat for Teaching – a free version of GenStat for educators and students. In today’s world of tightening budgets and justification of spending it’s vital that those at the forefront of teaching and coaching the next generation of scientists can access the best tools for in statistical analysis. With GenStat for Teaching now available free to all students and teachers world-wide, there is no reason why students cannot be taught using the best.

GenStat is known across the globe for its world class statistical tools and data analysis capability; from linear modelling to ANOVA and REML, in fact all that teaching requires. GenStat does, after all, stand for General Statistics! Its history and pedigree provides its users with reliability, trust and security.

On top of this, GenStat is one of the easiest data analysis packages on the market, with a clear and straightforward menu system to guide users through their analyses, backed up with dialogue boxes providing hints and alerts. A Save Session facility allows GenStat to be tailored around teaching sessions, breaking down the steps to fit with the learning path. A snapshot of a session can be taken at any time, and restarted at a later date, with no need to run through all the analyses again. These two benefits alone make GenStat an ideal tool for teaching; couple this with extensive statistical tests included in the software and you have a powerful tool to aid the teaching of statistics and data analysis techniques.

Additionally all users have access to the resources on our website with training available and tutorials for regression and ANOVA.

Download your free copy of GenStat for Teaching here .

New edition dispatch changes

All supported users should by now have received their CD for the GenStat 11th edition, and we hope you are enjoying and benefitting from the new developments.

In an attempt to reduce our carbon footprint, we are hoping to limit the amount we physically dispatch. As such we would like to make future upgrades available as download only from our secure website, and no-longer send out CD’s where possible. Should you still require a CD to be sent please could you email us, otherwise we will email details of any new upgrade and send you the download link.

We hope you see this as a positive step, as we do.

Technical tip – User Support

For more support and assistance don’t forget our on-line user guides for both GenStat and ASReml. These guides have also been updated and revised to include help on GenStat Discovery and GenStat for Teaching. The guides have everything from getting started to detailed statistical analysis, including reviews on the underlying methodology, explaining the output and describing the GenStat commands. Look at the full list of available documentation on our website.

Out and about with VSNi

A list of where VSNi will be this year is on our website.

On 21st July 2008 an enthusiastic group of GenStat users and developers gathered at the Agri-Food & Biosciences Institute (AFBI) in Belfast for the 14th European GenStat Applied Statistics Conference.

[conference attendees image]

The programme of 12 talks was split evenly between developments and applications, with topics ranging from forest fires and salmonella in Australia to sea birds in Scotland. More locally, Irish subjects involved tuberculosis in dairy cattle and sex ratio bias in gall midges. Other application areas included microarrays, microneurography, environmental assessments and the attraction of pollen beetles to oilseed rape flowers.

A broad range of statistical topics were covered, including generalized linear mixed models, hierarchical generalized nonlinear models and meta analysis, in addition to the more usual ANOVA, REML and regression analyses. New GenStat facilities were also described, with talks on survey analysis and the new facilities in the 11th Edition for canonical multivariate analyses, graphics environments and for using mathematical algorithms from the NAG Library.

We enjoyed excellent weather, and those staying on for the Advanced Linear Models workshop on the following day also enjoyed the excellent food, drink and ambiance of Belfast during the intervening evening.

We would like to thank AFBI for hosting the meeting, and especially the local organiser Alan Gordon (on the right in the picture below with Roger Payne).

[Roger Payne and Alan Gordon image]

On 5-9th October, Roger Payne is again speaking at the Joint Annual Meeting of the GSA, SSSA, ASA, CSSA, GCAGS and HGA in Houston, Texas. Roger will be presenting on A Guide to Analysing Counts and Proportions in Complex Situations, during a session on New Statistical Techniques for the Analysis of Agricultural Experiments. Roger’s talk will describe the types of biological investigation that have led to the development of methods such as generalized linear mixed models and hierarchical generalized linear models. It will show how these methods extend the more familiar generalized linear models to allow you to take account of additional sources of error variation, such as blocking in a field experiment, or parental effects in animal experiments.

If you are involved in organising an event which may be of interest to VSNi and our users please let us know by emailing us.

Latest training courses

We have an ASReml course on 24-25th September. This will be held by Dr Arthur Gilmour and will include an Introduction to Mixed Models and ASReml and on the second day Genetic Analyses for plants and animals. Participants are encouraged to bring their own examples, which can be sent direct to Dr Gilmour a month before the workshop for preparation. A similar workshop will also be taking place in Buenos Aires, Argentina 9-12th September to include the following areas:

– Introduction to Mixed Models
– Introduction to ASReml
– Spatial Analysis Theory and Practise
– Start OWN analyses
– Factor Analytic Model
– Repeated Measures
– Multi Environment trials
– Basic QTL Theory and practise
– Completion of own analyses

An applied workshop on Mixed Models for Plant Improvement using ASReml and r is planned for 2-5th November 2008, at the University of Western Australia, Perth. The workshop will present advanced statistical methods for the design and analysis of data arising from plant improvement programmes. Topics will include the design and analysis of single/multi environment and single/multi-phase experiments. Methods will also be presented for the integration of molecular marker and pedigree information into the analysis (and design) of these experiments.

As a part of our continued update and development of our courses, please let us know if you have any suggestions or topics for future training.

GenStat in Epidemiology

GenStat is well known and highly regarded throughout the world in its historical core area of biosciences, and specifically in agricultural research. The breadth of statistical analysis covered is well documented on websites, review articles and the like. As are the importance of its pedigree, developed, tried, tested and used by agricultural statisticians; the birthplace of GenStat (Rothamsted Experimental Station) being also the birth place of modern statistics with the likes of Sir Ronald Fisher, Frank Yates and Professor John Nelder all giving GenStat a certain kudos in statistical circles and the bioscientist’s world.

More and more disciplines are relying on statistics to uncover trends, causes and to better understand relationships between various factors. One area that has always understood the importance of statistics is epidemiology – the study of factors affecting the health and well-being of populations. Epidemiology is a vital discipline underpinning evidence-based medicine, for identifying risk factors for diseases and health effects.

The epidemiologist’s work ranges from investigations into disease outbreaks, clusters and exposure-response relationships, which may include the development of regression models to test hypotheses and estimate risk coefficients. The epidemiologist’s work at the Institute of Occupational Medicine in Edinburgh is designed to provide reliable information about health effects and risks for occupational and environmental hazards, with a view to addressing public and industry concerns, and providing a scientific basis for policies to limit disease. So it’s easy to see how a statistical analysis system such as GenStat is a vital tool for these researchers.

The IOM has been using GenStat for several decades in their studies on public health in the UK. Originally set up as a charity in 1969 to research coalminers’ lung disease, to continue a research programme set up by the National Coal Board’s medical service, the charity has been independent since 1990, and now provides research, consultancy, laboratory and measurement services in relation to potential health problems caused by occupational and environmental exposures. All the research reports published by the IOM since 1969 are available for free download from the on-line library.

GenStat has been used in a variety of different analyses, including epidemiological or observational data, which typically requires a regression model of some kind (linear, GLM, GAM, LMM, GLMM etc). It is also used for analysing data sets from designed toxicology experiments and for analysing cause-specific mortality data in comparison with reference rates.

A recent study looked at mortality rates in a group of almost 18,000 coalworkers from 10 collieries recruited from the 1950s onwards and followed up until the present time, of whom about two thirds are now deceased. One aim of the study was to compare the observed rates from certain causes of death with the male population rates for those causes in the regions where the coal pits are located. The calculations produce standardised mortality ratios (SMR’s) and their standard errors, using standard epidemiological methods.

GenStat used each individual’s entry and death or censoring dates to amass the person-years in the cohort, tabulating them by region, year and age (using GenStat’s option for sequential tabulation). The SMR calculations then used GenStat’s table manipulation functions to organise observed deaths and calculate expected numbers, ratio of observed to expected (SMR) and its standard error, etc. The outputs included overall SMR, plus a breakdown on 5 year-time groups that show how the healthy worker effect exists in the early part of the follow-up. The study has also been able to show that the risks of developing certain respiratory diseases increase with increased exposure to dust. Detailed results are available in a final report.

Cause of death

Observed deaths

SMR %

Confidence bounds

Lower

Upper

All causes

10698

100.9

99

102.9

All external causes

278

87.5

77.8

98.4

All internal causes

10421

103.7

101.7

105.7

 

 

 

 

Tuberculosis

16

77.8

47.6

126.9

 

 

 

 

All cancer

2732

98.0

94.4

101.8

Stomach Cancer

318

129.0

115.6

144.0

Lung Cancer

958

98.7

92.6

105.1

 

 

 

 

Cardiovascular Disease:

4890

97.8

95.1

100.6

Ischaemic Heart Disease

3298

100.2

96.8

103.7

Acute PHD

28

71.1

49.1

102.9

 

 

 

 

Non-Malignant Respiratory Disease

1966

138.2

132.3

144.5

COPD

849

115.5

108.0

123.6

Chronic Bronchitis

500

138.9

127.3

151.7

Emphysema

70

164.4

130.1

207.8

Pneumoconiosis

288

NA

NA

NA

CWP

222

NA

NA

NA

Silicosis

10

NA

NA

NA

The summary results of comparisons of mortality in cohort with external reference rates are shown below. The table shows, for chosen cause groups, numbers of deaths, age- year- and region-standardised mortality ratios (SMR) and 95% confidence interval.

[SMR for all internal causes]

The graph shows the Standardised Mortality Ratio (SMR) for all internal causes over the length of the follow-up period, with years grouped. The solid line is the SMR while the dashed lines represent the 95% confidence interval. The dotted line shows the SMR equal to 100%.

For any complex statistical calculations a software programme that is easy to use and reliable is crucial, but specifically in this instance GenStat’s table functions make the SMR calculations “beautifully simple to program.” (Dr Brian Miller).

The ability to understand the causes of health issues, what factors may lead to ill health or mortality in populations are of critical importance world-wide: so a sound, reliable data analysis system such as GenStat is vital to assist with analysis and help produce scientifically based recommendations and policies.

Our thanks to Dr Brian Miller of The Institute of Occupational Medicine for his help in producing this feature. More information on the IOM can be found here.

Images/Tables with permission from IOM research report TM/07/06.

[IOM Logo]

Discussion

Comments are disallowed for this post.

Comments are closed.