A NEW METHOD OF STATISTICAL ANALYSIS FOR SUGARCANE DISEASE SCREENING TRIALS

MANY ENDEMIC DISEASES are controlled through the cropping of resistant sugarcane varieties. Routine resistance screening trials for a range of diseases are a core activity of BSES. Central to this is the appropriate statistical analysis of data from these trials. This is important since potentially high-yielding varieties could be discarded from the breeding program if they are incorrectly rated as susceptible, or resources may be wasted on varieties that are incorrectly rated resistant. The worst situation is if varieties are incorrectly rated as resistant when released to industry and the end users then suffer serious losses. The current method used by BSES for rating varieties for disease resistance is based on a regression equation describing the relationship between the disease score of a set of standard varieties in the trial and the historical ratings of each standard. This regression equation is then used to assign a rating to the test varieties. In a recent review of the analysis of disease resistance trials, a new technique based on analysis with a linear mixed model was developed. The standard varieties are separated into groups based on the least significant difference test and the test varieties are placed in the group to which they best fit. A correlation is performed on the score of the standards in the new trial against the average score from all previous trials to determine if the current trial can be considered to provide a reliable estimate for the standards. In this paper, these new methods are applied to leaf scald and smut screening data and the impact on disease ratings and selection of new varieties is discussed.