When experimental constraints limit the randomization of the experimental order, it becomes necessary to perform analysis using a split-stage experimental model. When experimental data is neatly arranged, this has been handled by manually rearranging orthogonal arrays or complete ANOVA tables, although this is tedious. However, when the data is disorganized, this becomes impossible. JMP has sophisticated analytical functions in its "model fitting" section that can handle split-stage and branched experiments with disorganized data. This is a summary of the JMPer's Meeting held on March 2, 2006. Several "JMP Split-Stage Experiment Study Sessions" were held afterward, but they did not last long. In recent years, it is gratifying that JMP has become widely used in the manufacturing industry, and that JMP's DOE has attracted attention for experimental planning and analysis. In real-world experiments, there are always some constraints on randomization, but in practice, analyses are performed assuming complete randomness, and erroneous interpretations of results are common. Recognizing that the situation remains the same as it was 20 years ago, I would like to present this as an introductory course, including the posthumous writings of the late Professor Haga.
