TY - JOUR T1 - Use of Pretreatment Information in Design of Experiments. II. Nearly Optimum Designs for Homogeneous Groups JF - American Journal of Enology and Viticulture JO - Am J Enol Vitic. SP - 178 LP - 181 DO - 10.5344/ajev.1980.31.2.178 VL - 31 IS - 2 AU - Shu Geng AU - Michael B. Carter Y1 - 1980/01/01 UR - http://www.ajevonline.org/content/31/2/178.abstract N2 - Before experimental units are assigned to treatments, information may exist on the units that can be used to evaluate their uniformity. One objective in using this information is to allocate experimental units to treatments so that group means and group variance-covariances are as similar as possible, minimizing potential confounding effects on treatment responses. A procedure is proposed for constructing designs that have near homogeneity. One advantage of the approach is that no particular model need be specified at the time of allocation. Also, the procedure is invariant with respect to linear transformations and can identify outliers in the pretreatment data. An example is given of a vineyard study to assign sixteen plants to CO2 treatments in a photosynthesis study. A weighted procedure that incorporates sample sizes as weights for assignment is recommended for balanced designs. ER -