Comparing means of two categorical factors including interactions
Apr 10, 2018 9:11 PM(2498 views)
I am analyzing some data collected over the past year. For privacy reasons, I don't want to dive into it too much, but I think I've developed a simplified version of the experiment that I can use to ask questions.
I have a randomized design with a 3x4 factorial arrangement. We'll say that I have a "Day" factor (Day 1, Day 10, Day 30) and a "Treatment" factor (Approach A, Approach B, Approach A+B, and control Approach C). I am interested in the response variable "Value" (a numeric score indicating their performance on 100 point assessment). For each combination of Day and Treatment, let's say I have 3 participants. In other words, my data is something like:
30, 31, 33
50, 52, 55
90, 92, 93
32, 30, 31
40, 41, 43
75, 73, 76
I want to compare Value means between the Treatments and Days AND control for the interaction of Day*Treatment. So my question is, how do I do this in JMP? Am I using "Fit Model" or "Fit Y by X"? In the former, I don't see how I'm comparing means. In the latter, I don't see how to incorporate the interactions effect.
I apologize in advance - my speciality is Computer Science, not Statistics, so I might be a little slow on the uptake. I'm fine with a scripting answer, but I'd much prefer interactive explanation.
Even though I am not new to statistics, I always recommend plotting results. I reformatted the example data you supplied into a stacked format (often called the Anova format ). Attached the the table is a script to draw the variability plot.
Once you run it, put your mouse (pointer) over Day to the right of the horizontal axis, and drag Day to Treatment.
A significant interaction will be found with Fit Model, if the size of difference of the effect between A abd B is different by Day or the size of the Day effect is different by Treatment ( A vs B). Interpretting the calculated size of an effect and interactions is much easier using a graphical representation of the effects.