Comparison of methods for adjustment of plot yields with unequal stand. I. Simulated data

Abstract

A simulation study was made for comparing seven adjustment procedures of plot yields with unequal stand, namely: 1) analysis without adjustment; 2) adjustment proportional to initial stand; 3) based on Zuber's (1942) formula; 4) analysis of covariance; 5) analysis of covariance with extrapolation to initial stand; 6) analysis of covariance as proposed by Cruz (1971); 7) adjustment based on estimated compensation factor. Yield data of 10,800 plots of 180 experiments were simulated based on the following initial parameters: coefficients of variation of 8%, 12%, 16% and 20%; average final survival rates of 82%, 74% and 66%, in relation to initial stand; different compensation rates of yield increase due to missing plants in plots, varying from 0.1 to 0.9. Methods were applied to all 180 experiments. Efficiency of methods was measured by the sum of squares of deviations between adjusted treatment means and corresponding parametric means. Main conclusions drawn: a) method 5: most efficient under all situations resulting also in smaller experimental error; b) methods 6 and 7: also efficient under high experimental precision and high final survival rate; c) method 7: advantageous for providing reliable estimates of the capacity of cultivars in recovering yield lost due to reduced stand.