On the estimation of viabilities in competition experiments

Abstract

The biases in three currently used measures of relative viability are evaluated exactly and compared with previous approximate estimates of the bias. It is shown that the approximate estimate is often seriously misleading for one of the measures. The variances of the three estimators are also evaluated and it is shown that the bias is usually a lesser problem than the variability. The probabilities of the estimators being defined (i.e., having non-zero denominators) are also presented. Exact confidence limits are calculated and attention is drawn to their size. An alternative approximate method based on the log it is discussed. The experimental method of Anxolabehere et al. (1982) gives considerable estimation problems, particularly with low gene frequencies. The large variability of the estimators implies that only relatively large effects can be detected. © 1984 The Genetical Society of Great Britain.