Evaluation of biometrical methods for estimating the number of genes : 2. Effect of type I and type II statistical errors.

Computer simulation of several genetic models was used to assess the effect of type I and type II statistical errors on estimating the number of genes by the inbred-backcross and genotype assay procedures. Depending upon the actual number of genes, heritability, and the probability of type I errors (α), substantial upward and downward biases were observed in estimates of the number of genes from both methods. The estimated number of genes increased as α was increased from 0.01 to 0.30 and as heritability increased. With high α and/or high heritability, the estimated number of genes often exceeded the actual number. Downward biases occurred with low α and low heritability, and tended to become greater as the number of genes in the model was increased. Large type II errors were associated with downward biases. The choice of α had a greater impact on biases in estimates from the genotype assay procedure than from the inbred-backcross procedure. Increasing the number of backcrosses in the inbred-backcross procedure or delaying the assay generation in genotype assay increased the probability of upward biases in the estimated number of genes. Unbiased estimates can be obtained only by choice of an optimum α. There is no known way to choose the optimum α in practice. This fact reduces the value of estimates of the number of genes by genotype assay or by the inbred-backcross methods.