15 th European GenStat and ASReml Applied Statistics Conference Conference Abstracts Rothamsted Research July 2010

Conference Programme 0900-0930 Registration and Coffee 0930-1100 Session 1 Chair: Sue Welham 0930 Robin Thompson REML past, present and future 1000 Freedom Gumedze An alternative approach to outliers in meta-analysis 1030 Salvador Gezan Genetic analysis of tree clonal trials: combining individual and clonal information into ASReml Comparing statistical software for outputs on analysis of variance in agricultural experiments Julia Sibiya, John Derera & Pangirayi Tongoona Grain yield stability analysis of African maize hybrids in multi-environment trials based on the GGE biplot in GenStat Rodger White & Sarah Kemmitt Phospholipid fatty acids and analysis of distance for compositional data There is a need to non-invasively and reliably detect whether drugs aimed at targets in the Central Nervous System (CNS) cross the blood/brain interface and have an effect. However it is not possible to directly assay drug levels in the CNS, so indirect methods are needed. One of these is the electroencephalogram (EEG), which provides a non-invasive method of detecting brain activity. Typically time slices of the EEG signal are subject to a Fourier Transform which gives a periodogram of the signal versus frequency at a sequence of time points. The SLCCA algorithm (Gregg, Varvel & Shafer 1992) is capable of detecting a frequency "signature" in these periodograms which is correlated with the imputed dose level of a drug in the brain. The approach uses Canonical Correlation Analysis to detect the signature which maximises the correlation between the EEG periodogram and the imputed dose obtained from a PK model with unknown nonlinear parameters, where the unknown nonlinear parameters of the PK model are estimated by maximising the Canonical Correlation. We have extended the previous algorithm to include regression models with linear and nonlinear parameters so that a range of complex models can be readily fitted. We intend to make the algorithm generally available as a GenStat procedure in due course. We will present the algorithm and illustrate its use in a regression context on experimental EEG data from rats. 1. Background Many insects are known to migrate long distances to exploit seasonally available habitats. The painted lady butterfly (Vanessa cardui) over-winters in North Africa and migrates to northern Europe in spring. The use of the sun in orientation has been demonstrated in various insects, including bees, beetles and ants. 2. The scientific questions Do painted lady butterflies a) preferentially orientate northwards during spring migrations? b) use a sun compass to select and …