Time: Refreshments from 16:45 for a 17:00 start (duration of seminar 1h).
Non-members welcome (as always).
Location: University of Queensland, St Lucia Campus, S524, Hartley Teakle, Building No. 83, (level 5 South)
Speaker: Dr. Sue Welham, VSN International Ltd, Hemel Hempstead, UK
Title: Finding the optimum: interaction between statistical and logistical issues in dose-response field trials.
Authors: Sue Welham (VSN International), Roger Sylvester-Bradley, Daniel Kindred, Liz Hudson, Susie Roques (ADAS UK Ltd)
Standard variety trials measure yield and quality traits for a range of varieties in the field, with some level of replication, using a standard dose of nitrogen fertiliser. To achieve “sustainable intensification”, we want to identify varieties that give maximum yield for minimum input. This requires dose-response trials to identify optimum doses for varieties, by testing at several doses of standard inputs (here nitrogen), which greatly increases the number of plots and trialling costs. In addition, the large plot-to-plot variation associated with field trials can make small differences in optima hard to detect. Taken together, these factors make wide-scale testing of dose-response curves impractical and so varieties with efficient nitrogen-use properties may never be identified.
Use of precision farming equipment has increased the feasibility of small-plot experimentation and ADAS UK introduced the concept of OptiPlot trials: a range of doses applied in increasing order within a standard field plot, with varieties replicated across field plots. The dose-response trial can then be undertaken on the same scale as a standard variety trial, with the added benefit that dose comparisons take place in close proximity and should be more precise. As originally defined, this concept ignores the statistical principle of randomization, and gives the possibility of confounding between any underlying spatial trend and the nitrogen response.
This talk will describe the evolution from the original OptiPlot concept into one that provides a practical compromise between good statistical practice and logistical constraints. The role and limitations of simulation studies in the evaluation of candidate designs will also be discussed.
This work forms part of the Agronomics Project funded by Innovate UK, led by ADAS UK Limited and involving the Courtyard Partnership, BASF, Trials Equipment Ltd., VSN International and the British Geological Survey.
Biography: Sue Welham is an applied statistician, currently working on algorithms for linear mixed models at VSN International (UK). She started her career at Rothamsted Research, working as a consultant statistician with Rothamsted scientists and writing a front-end to the REML algorithm in GenStat. She did a PhD on mixed model splines at the London School of Hygiene and Tropical Medicine during 2000-2003 then returned to Rothamsted to head the group of statistical consultants. During this time, the group developed a training program in practical statistics for scientists that led to the publication of “Statistical Methods in Biology” by Welham, Gezan, Clark and Mead (2014). Sue moved to VSN in 2012 to resume her interest in computational algorithms for REML estimation, and also retains some collaborations with scientists at Rothamsted and ADAS.
|Time:||4:45 pm - 6:00 pm|