Methodology for Estimation of Empty Cells
with the Mean Cell Model in Connected Designs
Sandra Melo, Luis López & Oscar Melo
In this paper we propose two methods to estimate empty cells and to impute
missing data in connected designs. The first is supported on Bartlett’s covariance method on the basis of
cell means model with effect restriction; the second is based on the least
squares estimation method with no interaction. A general expression is proposed
to estimate cells with missing information. The equality of the estimators
obtained by these methods is demonstrated and a lower bound to the number of
cells needed to be estimated, so that the design is connected, is calculated.
The adjusted anova is presented and finally, a 3-way no interaction fixed
effects model is used to illustrate the two methods.
Key words: Connected designs, Mean cell models,
Estimable functions, Restrictive model, Unbalanced designs.