Optimum plot size and number of replications with short-day onions for yield, seedstem formation, number of doubles, and incidence of foliar diseases

Abstract

Five different statistical methods were used to estimate optimum plot size and three different methods were used to estimate optimum number of replications with short-day onions (Allium cepa L.) for yield, seedstem formation (bolting), purple blotch and/or Stemphylium (PB/S), botrytis leaf blight (BLB), and bulb doubling with a basic plot size unit of 1.5 x 1.8 m (length x width). Methods included Bartlett's test for homogeneity of variance, computed LSD values, maximum curvature of coefficient of variation plotted against plot size, Hatheway's method for a true mean difference, and Cochran and Cox's method for detecting a percent mean difference. Bartlett's chi-square was better at determining optimum plot size with transformed count and percent data compared with yield data in these experiments. Optimum plot size for yield of five basic units (7.5 m length) and four replications is indicated using computed LSD values where the LSD is <5% of the average for that plot size, which was the case in both years of this study. Based on all the methods used for yield, a plot size of four to five basic units and three to five replications is appropriate. For seedstems using computed LSD values, an optimum plot size of four basic units (6 m length) and two replications is indicated. For PB/S two basic units (3 m length) plot size with four replications is indicated by computed LSD values. For BLB a plot size of four basic units (6 m length) and three replications is optimum based on computed LSD values. Optimum plot size and number of replications for estimating bulb doubling was four basic units (6 m length) and two replications with 'Southern Belle', a cultivar with a high incidence of doubling using computed LSD values. With 'Sweet Vidalia', a cultivar with low incidence of bulb doubling, a plot size of four basic units (6 m length) and five replications is recommended by computed LSD values. Visualizing maximum curvature between coefficient of variation and plot size suggests plot sizes of seven to eight basic units (10.5 to 12 m length) for yield, 10 basic units (15 m length) for seedstems, five basic units (7.5 m length) for PB/S and BLB, five basic units (7.5 m length) for 'Southern Belle' doubling, and 10 basic units (15 m length) for 'Sweet Vidalia' doubling. A number of plot size-replication combinations were optimum for the parameters tested with Hatheway's and Cochran and Cox's methods. Cochran and Cox's method generally indicated a smaller plot size and number of replications compared to Hatheway's method regardless of the parameter under consideration. Overall, both Hatheway's method and computed LSD values appear to give reasonable results regardless of data (i.e., yield, seedstems, diseases etc.) Finally, it should be noted that the size of the initial basic unit will have a strong influence on the appropriate plot size.