The Use of Probits in Combining Percentage Kills

The problem is discussed of how to combine percentage kills obtained in an experiment repeated under different sets of conditions. Treatment effects and their interaction with replicates are defined in terms of probits, and, assuming the absence of such interaction, a method is evolved of finding maximum likelihood estimates of the mean treatment differences. An example is used to illustrate details of the method. The merits of the logit and angular transformations applied to the same type of data are also discussed.