Lecture 30: More on Factorial Crossing Treatments and Factors Example 1 Factor A: two levels Factor B: two levels The four treatment combinations are replicated 5 times. Completely randomized design. You can’t tell if factors are statistically significant even if treatments are statistically significant without doing the appropriate tests. 1 2 Treatments Factors There are statistically significant treatment effects. F = 4.2244, P-value = 0.0222. Treatment A1,B2 has the highest mean and it is statistically different from all other treatments. No factor nor interaction is statistically significant. F values of 4.224, P-values of 0.0565. 3 Example 2 4 Treatments Factor A: two levels Factor B: two levels The four treatment combinations are replicated 5 times. Completely randomized design. There are no statistically significant treatment effects. F = 2.1331, P-value = 0.1361. 5 6 1 Lecture 30: More on Factorial Crossing Factors The interaction between Factors A and B is statistically significant. F value of 6.3992, P-values of 0.0223. 7 Handout The effect of Factor B is positive for Level A1 but negative for Level A2. This is indicative of an interaction between the two factors. 8 Parameter Estimates JMP produces Parameter Estimates as part of its Fit Model output. These are related to estimated effects. See the handout on the course web site for the data and JMP output for both examples. 9 Dental Pain Relief 10 Estimated Drug Effects Parameter Estimates Term Estimate Intercept 1.15625 Drug[Codeine] 0.26875 Acupuncture[Active] 0.32500 Drug[Codeine]*Acupuncture[Active] 0.0375 11 Level Codeine Placebo Mean Estimated Effect 1.4250 1.4250 – 1.15625 = +0.26875 0.8875 0.8875 – 1.15625 = –0.26875 The parameter estimate for Drug[Codeine] is the estimated drug effect for Codeine. 12 2 Lecture 30: More on Factorial Crossing Estimated Acupuncture Effects Level Active Inactive Estimated Interaction Effects Active +0.0375 –0.0375 Codeine Placebo Mean Estimated Effect 1.48125 1.48125 – 1.15625 = +0.325 0.83125 0.83125 – 1.15625 = –0.325 The parameter estimate for Acupuncture[Active] is the estimated acupuncture effect for Active. Inactive –0.0375 +0.0375 The parameter estimate for Drug[Codeine]*Acupuncture[Active] is the estimated interaction effect for Codeine*Active. 13 14 Model Estimated Treatment Effects JMP does not provide a parameter estimate that gives the estimated treatment effects but you can calculate these from the parameter estimates given. 15 Parameter Estimates 16 Estimated Treatment Effects Drug[Codeine] + Acupuncture[Active] + Drug[Codeine]*Acupuncture [Active] = 0.26875 + 0.325 + 0.0375 = 0.63125 17 Level Codeine,Active Codeine,Inactive Placebo,Active Placebo,Inactive Mean 1.7875 1.0625 1.1750 0.6000 Estimated Effect 1.7875 – 1.15625 = +0.63125 1.0625 – 1.15625 = –0.09375 1.1750 – 1.15625 = +0.01875 0.6000 – 1.15625 = –0.55625 18 3