Validity of f-test in a randomized block design with missing data

摘要: In this paper the first two moments of test criterion (Treat. S.S.)/(Treat.S.S. + Error S.S.) have been derived assuming that one observation corresponding to plot block which is under treatment ‘m’ (say), missing in a randomized design with ‘V’ treatments and ‘r’ blocks. To keep analysis simple case has considered. It concluded general not unbiased YATES (1951) sense usual F-test satisfactory iff /(vr - r 1)=(S S1)/(v 1) (r-1), errors are homogeneous (vr large. The three unifornity trial data indicates condition most important for be satisfactory. However, if at random from some block, unbiased. If GT96, also provides good approximation randomization case.