Vision No. 10 Oct 2008

[apples]

As a leading supplier to agricultural research it’s fascinating to watch the stir being created by articles on GM crops and the use of pesticides, especially with the proposed EU legislation. But it’s not necessarily clear which route governments and the industry should be taking…

From fears of poisoning the earth with misuse of pesticides versus the potential reduction in Europe’s food production and food price increases if some pesticides are banned, to the ethical and moral issues over genetically modifying crops at one end of the spectrum to the real need and ability to provide crops that can withstand pests, diseases, droughts or other climatic conditions. And now recent research suggests that some GM crops may even protect neighbouring non GM crops (American
Association for the Advancement of Science (2008 September 19) “Genetically Modified Crops Protect Neighbors from Pests”). Each side of the argument seems compelling, so what should we do?

One thing is clear, sound reliable research is required for sensible future planning. And that is something we at VSNi can help with. By providing outstanding data analysis software, specialising in the biosciences, throughout the world we can assist researchers in each area to come up with results and projections based on solid data analysis.

VSNi software is firmly rooted in the biosciences and agriculture. GenStat, originally developed by statisticians at Rothamsted Research in the UK, is still extensively used there today. Projects range from developing strategies to reduce the use of pesticides, improving production and quality of crops or by using the earth’s natural biodiversity to help improve production or crop yields and
reduce the need for pesticides.

Improving yields covers the simple definition of “more” to the quality of the crop and the ability of the crop to adapt to different environments. For example, researchers in a plant breeding department work on developing hybrids, which then need testing in large multi-year, multi-locations trials. The purpose of these trials is to select the hybrids which have the highest yield potential, and the best adaptation to particular environments. Here it’s clear that whilst crop yields can be
improved, there is very little purpose in doing so if they do not adapt well to their targeted environments. In these analyses GenStat’s REML has been particularly useful for spatial analysis of trials conducted on uneven soils and for meta-analysis of large trial datasets; this allows researchers to evaluate genetic gain and the general and specific combining ability. The crucial issue in this
instance is the ability of the chosen data analysis package to provide complex statistical approaches, in an easy to use way; i.e. experimental design, analysis of variance and mixed model approaches (REML), but in a menu system.

“The experimental design tools of GenStat are really excellent and user-friendly. I think that it is the only statistical program that offers complex statistical approaches that are extremely useful in agricultural research in a friendly, easy-to-use way.”Abelardo de la Vega, Advanta Semillas

Another aspect of increasing yields is how herbicides are used. Researchers in Latvia are testing for the effectiveness of different herbicides to control weeds and maintain crop safety. One such research project centred around surveys on flora in five arable fields in five regions of Latvia aims to assess the effects of crop rotation and crop husbandry practices. These projects are aimed at providing scientifically based information to help agronomists and farmers determine the need for
weed control.

GenStat’s REML techniques were key as the number of contributing factors varied and using this technique allows for the analysis of unbalanced datasets, and produces output equivalent to the analysis of variance. Researchers were able to include over sixty species of weed, and analyse the effect of the use of different herbicides, on different crops including the effect previous crops have on the
trials.

Another customer of VSNi’s is The Organic Research Centre at Elm Farm in the UK. The centre is designed to look at providing solutions that develop and support sustainable agriculture and land use. Based upon organic principles to ensure the environment’s health is as protected as possible, the research programs are conducted at the farm in Berkshire and across 25 other farms in the
UK. GenStat is primarily used in the wheat breeding trials, looking at how Composite Cross Populations (rather than specific varieties) perform under fluctuating environmental conditions.

“The great aspect of this new analysis is that we can use data from all 12 experiments, i.e. 3 years over 4 sites, to work out which of the varieties and populations are both yielding and reliable,” says Sarah Clarke, ORC, “we can also split the experiments into those that are organic and those that are non-organic, to see if the populations differ between systems.”

The aim is to produce wheat that performs well year after year in differing environments. It is therefore vital that the tools used by these researchers are reliable and provide sound statistical analysis techniques, so that the results from the trials are as trustworthy as possible.

ASReml is another such trusted product. Again born out of agricultural research requirements, this time in the analysis of 12 years of data from over 1000 wheat variety trials, today ASReml is used across the world by plant and animal breeders to help solve some of the big puzzles found in these areas. Its ability to handle the large datasets so often found in agricultural research is one of its main
strengths; indeed the National Variety Trials in Australia, handling data from around 600 trials across Australia in over 250 distinct geographical locations, rely on the results from their trials using ASReml to give reliable predictions on genetic value of different crops in different locations.

“ASReml is the package that helps farmers, breeders and crop variety evaluators obtain the most reliable predictions of genetic value for a range of crops grown in different environments; farmers can get the best information available about performance of varieties in their own location and make an informed decision,” says Professor Brian Cullis, Research
Leader for DPI Biometrics and leader of the SAGI (Statistics for the Australian Grains Industry) project.

Another key strength of ASReml is the accuracy, speed and flexibility of the complex two-stage linear models used in these trials.

But it isn’t just the big Western companies that benefit from our software. Since 2003 we have provided GenStat Discovery, (now at edition 3) to research and teaching institutes across the developing world. As Stewart Andrews, VSNi CEO says “there is no justification for poorer countries to not have access to high quality research tools. Their need for research based on solid,
reliable data analysis tools is as great, if not greater than the West’s; assistance to these countries should come in the form of providing opportunities for self help and self determination, not just aid or gifts.”

GenStat Discovery has become a key tool in the researchers kit bag; relied on by many agricultural institutes throughout the developing world; many of these researchers would not otherwise have access to data analysis software, nor would they be able to share their results and ideas with colleagues around the world.

“I have spent money on other software packages that fall short of my needs.” Ugwu Kenneth Okonkwo, University of Nigeria, Nsukka

For the CGIAR centres this is important, as they work closely with institutes in the developing world such as ILRI, ICRISAT or ICRAF, and need to be able to collaborate on research activities. It is the concept of trust that is so important in these types of trials and research projects; something that VSNi prides itself on. Our software is and has been trusted by
researchers for over 30 years. Trusted because of our history in agriculture – we understand the types of data and the analysis needs within this industry and trusted because of the sound statistical principles on which the software is based.

Technical tip – User Support

Felix Grant’s recent review of GenStat 11th edition mentioned many new features, including the ability to carry out partial or full canonical correspondence analysis (CCA) using a menu based system. Included on the menu is an option to plot a biplot of the results, showing how site or species scores are related to one (or two) of the environmental variables.

 

[CCA ordination biplot image]

Another option could be to run the command CRTRIPLOT to see a plot of species scores, site scores and biplot scores of the environmental variables in a single plot. The user has many options including being able to plot the scores in a three-dimensional environment, add convex hulls and other graphics for grouped data and set the dimensions to be plotted. See the help file for a full list of options.

Read Felix’s full review of GenStat 11th edition, or download your trial copy of the 11th edition.

Out and about with VSNi

As we enter the last quarter of the year, the travel doesn’t stop for VSNi. A key event we are looking forward to is the Australasian GenStat User Conference, from 2-5th December at Marysville, Victoria. A packed agenda on Biometrics in Primary Industries and the Environment, and with many of our developers attending, it’s your chance to share ideas and quiz us face to face. More information and to register is available at their website.

Roger Payne once again attended the Joint Annual Meeting of the GSA, SSSA, ASA, CSSA, GCAGS, and HGS, this year, held in Houston on 5 – 9th October; this year speaking on “A Guide to Analysing Counts and Proportions in Complex Situations”, and specfically covering the methods of analysing counts and proportions from the experiments often found in agriculture and biology. It described the types of biological investigation that have led to the development of methods such as generalized linear
mixed models and hierarchical generalized linear models. His talk also showed how these methods extend the more familiar generalized liner models to allow you to take account of additional sources of error variation. For wrap up information on this event go to the event website.

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

Latest training courses

The next training course is for REML and multi-level models in GenStat, scheduled for 14th November at The Paper Trail in Apsley, UK. To find out more or to book please look at the training page on our website.

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.

 

Correction

In the last Newsletter we focussed on Dr Brian Miller, of the IOM, and his work in epidemiology; VSNi would like to make a correction to the report number listed, which should have read:IOM research report TM/07/06. Read the article again here.

Discussion

Comments are disallowed for this post.

Comments are closed.