Vision No. 7 March 2008

New Office opens in Australia

VSNi is proud to announce the opening of a new office, based in Australia to support our customers in Oceania. More information is available on our home pages.

Masterminding REML

[VSNi logo]

Some years ago, on a remote farm in deepest, darkest Lancashire was born a bouncing baby boy named Robin. An unfavourable encounter with an over-friendly cow at an early age didn’t deter the young Robin from an ongoing affinity with animals and the countryside. This affinity combined with an exceptional appreciation of figures eventually led to a very useful statistical advance.In 1971 Robin Thompson and Des Patterson co-authored a paper on the methodology of residual maximum likelihood that extended analysis of variance ideas to the analysis of unbalanced correlated data; and removed the bias in estimation of variance components present in maximum likelihood methods.Whilst on a long haul flight to Australia, before planes doubled as entertainment complexes, the lead in our story sat patiently scribbling what to the untrained eye could only be described as cryptic etchings. Young Robin however was not an early advocate of Dan Brown but toying with the conundrum of how to conduct variety trials with large unbalanced data sets. A flash of insight, a splash of red wine and the Average Information algorithm for REML estimation was born; the algorithm that underpins the mixed model analysis in today’s GenStat and ASReml programs. When asking about his inspiration for such a revolutionary find Robin modestly says “I guess I’m just good with numbers.”

The ongoing achievements of our Dr Robin Thompson, the supreme advancements of the REML methodology and the recognition of all parties involved simply couldn’t be contained in one editorial. Some intriguing chapters, if space were available, would include imperfectly worded application forms, complex travel budgets and the christening of a Spanish Plaza. But we must move swiftly on to the esteemed collaborators, without whom the evolution and accessibility of the REML/ AI algorithm would not have been possible.

Here we recognize the brilliance of Dr Arthur Gilmour and Dr Brian Cullis; who joined forces with Robin in the early 90s. By 1995, with the dawn of larger variety trials and ever expanding data, Brian, Arthur and Robin pioneered techniques that would eventually evolve into ASReml. Over the years ASReml has revolutionized the way we consider problems providing a platform to analyse huge and complex data sets. Today it is used across the globe providing a rich and flexible solution for the analysis of data sets commonly arising in the agricultural, biological, medical and environmental sciences.The ASReml user guide has been cited in over 800 refereed publications.

Despite his recent retirement Robin still spends a great chunk of his time at RRES – when I questioned as to why…. “ASReml 3 my dear” was his ever dedicated reply. ASReml 3 is being developed and we can look forward to many new features including the analysis of grouped data and extra flexibility for defining genetic relationships particularly for inbred varieties.

More information on the current version of ASReml is available here.

Technical tip – User Support

There are two highly useful and often overlooked sections of the website – the User Areas for GenStat and ASReml. They collate resources produced by GenStatters and ASRemlers for you.

We always welcome further additions such as any course notes or libraries you might produce that others could find useful. Please e-mail support with any contributions.

The GenStat area has a link to the vast resource of the GenStat discussion list with over a decade of searchable archives – you can put your questions to the most dedicated, helpful and experienced users of GenStat. If you’re not looking for specific help but are starting out then the kind people of Reading University and their partners have put together various course notes and an introductory guide. Should you want to look at extending what GenStat can do then the researchers from the Biometris department of Wageningen in Holland have put together a procedure library with nearly 50 new procedures for you to use.

The ASReml area contains the wonderful Cookbook by Luis Apiolaza to get you started, with a host of useful ASReml recipes and some real food recipes mixed in for good measure. If genetic analysis is to your taste, then the paper (pdf) by Julius van der Werf shouldn’t be missed. Lastly we have a whole host of user contributed tutorials from getting started to advanced modelling.

We would like to thank all the people who contributed the areas so far which form very much the life blood that keeps GenStat and ASReml successful. Please pass anything that you think might be useful onto us and we will put it up for others to use.

Out and about with VSNi

Our events list is beginning to fill up for 2008, with meetings throughout the world.Don’t forget to register for the European GenStat Applied Statistics Conference at the Agri-Food & Biosciences Institute (AFBI), in Belfast, Northern Ireland, on 21st July 2008, (which takes place immediately after the International Biometrics Conference). It will be followed by a workshop on Advanced Linear Models on Tuesday 22nd July (also at AFBI) for more details see our website.

For those wanting more on HGLM’s John Nelder (Imperial College), Mike Kenward (LSHTM) and Roger Payne (VSNi) will be speaking at an RSS half day meeting on “Hierarchical Generalized Models and Beyond: Theory and Implementation” on 3rd April 2008, for more information or to register look here.

To see where VSNi will be in the future look here.

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

The latest training courses – An Introduction to GenStat (25th February) and Regression, nonlinear and generalized linear models in GenStat (26-27th February) have just been held in Hemel Hempstead. More courses will be announced soon on the VSNi website, if you have any suggestions for future courses please do contact the VSNi training team.

VSNi photography competition

Congratulations to our winner, VP Prasanth from ICRISAT in India, with this beautiful picture of a uniform sprinkler entitled “A big leap towards Green from Grey”. Our winner receives £150 of Amazon vouchers.[A big leap towards Green from Grey, by VP Prasanth, ICRISAT]To see some of the other entries please look here. Once again thank you to all those who entered, we really enjoyed seeing so much of where you work and what you do.

Statistics and animal welfare

[A donkey with lesions on the spine]

This donkey is seriously suffering with lesions on the spine. Obviously it must be the pack saddle that is causing the sores so let’s add extra padding under its pack saddle to make it better. If we’re lucky this might reduce the problem, but what if the problem is actually more complicated than that? What if the padding is important, but only when coupled with being dirty, or with rough handling from the owner, or a particularly awkward load? Then increasing the padding would be unlikely to make much impact, and besides, what type and thickness of padding would it be best to change it to? Many such welfare issues are faced by working equines in developing countries and the fact that they may be caused not by one obvious factor, but by many factors which are not always obvious and may interact with each other, is the underlying principle of the Risk Assessment project developed by the Brooke.

Historically animal welfare work abroad, has typically assumed the causes of such problems, but not necessarily known them or had evidence to support them. Hence treatment or prevention interventions may have been based more on assumption than fact.

However the Brooke takes a different approach. The Brooke is the UK’s leading overseas equine welfare charity, committed to improving the welfare of horses and donkeys in some of the poorest countries in the world. Countries where families are dependant on their donkeys, horses or mules for their work, and therefore the welfare of these animals can mean life or death for the family. And yet, many suffer through malnutrition, disease and injury due to poverty, misguided intention or lack of accessible or affordable veterinary care. Since 1934, the Brooke has been providing free treatment and training through their teams of mobile vets and animal welfare workers. They have been working hard to underpin their work with scientific evidence.

The Brooke believes in sustainable support. That is they look to finding long-term solutions to the problems and not just fixing them in the short term. To fully achieve this they need to understand what these welfare issues are and what causes them. This scientifically based approach is unique to the Brooke amongst animal welfare NGO’s abroad.

A Risk Assessment Team, headed up by Amanda Childs, studies welfare issues individually to identify and understand the root causes of the problem. Her team visit the communities and observe, measure and record data on all sorts of issues surrounding the problem being assessed. This data is then collated and analysed using GenStat to understand which are the key factors contributing to the welfare issue. The long term aim being to eliminate these welfare issues by designing specifically targeted interventions.

[A donkey with lip lesions]

For example, a study in Mardan and Gujranwala in Pakistan found that 70% of the horses, mules and donkeys pulling carts had lip lesions. The animals were in considerable pain from the lesions alone, let alone any secondary effects such as lack of ability to eat adequate feed.

The assessments are always incredibly detailed, and this one in Pakistan was no exception: 203 measurements and observations were made on more than 370 animals in a month. The data included many characteristics of the bit in the animal’s mouth, from its overall design to its rustiness, size and cleanliness. Close observations of the driving behaviour were made, such as the frequency with which the reins were pulled, any beating and the style of loading, as well as many other measurements taken from the working environment, the cart and bridle, and of course the animal itself.

The Brooke Risk Assessment staff has been training in statistics and the use of Genstat for the last two years. Their knowledge of Genstat not only allows them to present their work graphically and descriptively, but to analyse the relationships between the welfare issue (outcome) variables and explanatory variables, predominantly using linear and logistic regression analyses. The ’screening test’ facility enables them to explore both univariate and multivariate relationships far more quickly than with any other software, while the ability to calculate both conditional and marginal tests for linear models makes it quick and easy to identify potentially confounded variables.

The results of the lip lesion assessment were fascinating and showed that 20 different factors contributed to the lesions. Specifically these factors could be split into 4 groups:

1. the bit

2. the attitude and behaviour of the owner

3. the animal itself

4. the cart and harness.

The detail of the results mean specific advice and help can be offered to the animal’s owner, who depend on these animals for their livelihoods. Information about the type of bit to use, the use of a nose band and the best ways to drive the animals…in all 17 recommendations have been made by the risk assessment team, which can be used in the animal owning communities to dramatically reduce if not eradicate the problem of lip lesions on these animals.

The ease of use of GenStat’s menu system has meant that the Brooke’s field staff can now produce their own statistical analysis of the issues and are increasingly keen to do so, despite not being statisticians by training. Indeed most of the staff has veterinary or bioscience backgrounds. Therefore a system which provided world class statistics at the drop of a menu was vital. The aim for all these staff is that with GenStat all the Brooke’s field centres will be independent and able to perform their own analysis, where previously the data was sent to UK consultants.

“When choosing a statistical programme we presented 3 different programmes to our field workers and let them decide which one they felt would be most suitable. The final decision, based on the ease with which data could be manipulated and the ability to clearly see how the programme would be used in our specific context, was unanimously in favour of GenStat.” Amanda Childs, Head of Risk Assessment, the Brooke.

GenStat has made an enormous difference to the Brooke’s ability to easily and quickly analyse their own data and in Amanda’s words has provided the opportunity to “bridge a huge gap” between the need to provide quality data analysis and the need to get that done in a way that is locally sustainable.

Our thanks to the Brooke for their help with this article. All photographs in this article are courtesy of the Brooke.

More information on the Brooke.


Comments are disallowed for this post.

Comments are closed.