Main content area

A biometrical analysis of yield trials with timothy varieties using rod rows

Smith, H.F., Myers, C.H.
Agronomy journal 1934 v.26 no.2 pp. 117-128
Phleum pratense, variety trials, statistical analysis, crop yield, strain differences, row spacing, harvest date, equations, correlation, sowing, soil heterogeneity, mathematics and statistics, New York
1. The paper reports the results of a varietal yield trial with 39 varieties of timothy sown in September 1923 and harvested in 3 years, 1925-27. Plats were in the form of rod rows. The yields obtained are given in Table 12. 2. With the exception of three varieties obtained from Aberystwyth, Wales, the relative yields of varieties were the same, within the limits of experimental error, in all 3 years. The varieties gave the greatest yields in 1925 and the lowest in subsequent years. 3. Twelve of the 39 varieties were also grown in two other trials sown in 1920; one used rod rows and was harvested over 5 years, 1921-25; the other used broadcast plats, 16 1/2 feet by 49 1/2 feet, and was harvested in 3 years, 1921-23. It has been shown that the yields of these 12 varieties agreed in all three trials as closely as they could be expected to do having regard to the experimental errors of the trials. 4. Rod rows showed least variability in the second and third harvest years. In these years the standard error for a single row was about 9%. It is recommended that in rod-row trials of timothy only the yields in these 2 years should be recorded, unless it be desired to investigate ability to maintain yields in subsequent years. The error observed for the broadcast plats was 8% in the first year and about 20% in the second and third years. 5. The rod-row trial sown in 1923 had 10 replications of each variety, but these were arranged in a systematic manner. The trial presents a case in which the effects of soil heterogeneity were extremely marked because the long narrow plats (rod rows) lay along, instead of across, the fertility gradient, and in which, because of the systematic arrangement of replications, the mean yields of varieties were not directly comparable. The position was, however, retrievable by the use of check plats which occurred in every fifth row. Reasons have been presented by which it was deemed permissible to apply Fisher's analysis of variance to yields adjusted by relation to the checks. 6. The relation to parts in the analysis of variance of older statistics which have been used to estimate the errors of experiments extending over several years has been indicated.