ABSTRACT
The individuals are classified by three factors, water softness at three levels, temperature at two levels, and a factor whose two levels correspond to previous experience and no previous experience with standard product (M). The object is to study how preferences for new product (X) vary. The proportions of preferences in the different cells are not of equal precision. For fitting a saturated model, that is one with as many parameters as cells, these changes of variance are unimportant. For fitting reduced models, for estimating certain contrasts assuming others to be zero, revised estimates would be obtained by allowing for the changes of variance, fitting by maximum likelihood or weighted least squares. To investigate the precision with which the effects are established, the standard factorial contrasts have been calculated from the proportions. For the latter the proportions preferring X are nearly all below 12, are lower at the higher temperature and lower for hard water than for soft water.
