Genizi, A., Agricultural Research Organization, Bet-Dagan, Israel

Lahav, E., Agricultural Research Organization, Bet-Dagan, Israel

Putter, J., Agricultural Research Organization, Bet-Dagan, Israel

Lahav, E., Agricultural Research Organization, Bet-Dagan, Israel

Putter, J., Agricultural Research Organization, Bet-Dagan, Israel

The determination of optimum unit size requires the knowledge of the between-units variance Vg as a function of the unit size g. The problem is to estimate Vg from the data of an actual experiment (as opposed to a uniformity trial), in which different treatments were assigned to different units. Some estimators have been proposed in the literature, but they are based on models with underlying independence assumptions. The randomization model makes it possible to derive unbiased estimators for Vg without such assumptions. Some explicit formulae for such estimators, most of which are quadratic functions of the treatment residuals, are obtained for the completely randomized, randomized block, and balanced incomplete block designs. © 1984.

Estimation of the variance between units of given size from experimental data

9

Genizi, A., Agricultural Research Organization, Bet-Dagan, Israel

Lahav, E., Agricultural Research Organization, Bet-Dagan, Israel

Putter, J., Agricultural Research Organization, Bet-Dagan, Israel

Lahav, E., Agricultural Research Organization, Bet-Dagan, Israel

Putter, J., Agricultural Research Organization, Bet-Dagan, Israel

Estimation of the variance between units of given size from experimental data

The determination of optimum unit size requires the knowledge of the between-units variance Vg as a function of the unit size g. The problem is to estimate Vg from the data of an actual experiment (as opposed to a uniformity trial), in which different treatments were assigned to different units. Some estimators have been proposed in the literature, but they are based on models with underlying independence assumptions. The randomization model makes it possible to derive unbiased estimators for Vg without such assumptions. Some explicit formulae for such estimators, most of which are quadratic functions of the treatment residuals, are obtained for the completely randomized, randomized block, and balanced incomplete block designs. © 1984.

Scientific Publication

You may also be interested in