Experimental Design Answer

All data analysis performed on SPSS, a joint venture partner with LWI.

Tests of Between-Subjects Effects

 

Dependent Variable: Experimental Output

Source Type III Sum of Squares df Mean Square F Sig. Eta Squared
Corrected Model 104.725a 3 34.908 9.773 .000 .449
Intercept 9620.777 1 9620.777 2693.542 .000 .987
Setting A 40.802 1 40.802 11.424 .002 .241
Setting B 1.238 1 1.238 .347 .560 .010
Setting A * Setting B 62.684 1 62.684 17.550 .000 .328
Error 128.585 36 3.572      
Total 9854.086 40        
Corrected Total 233.309 39        

a R Squared = .449 (Adjusted R Squared = .403)

To minimize Output, run Setting A at 20 and Setting B at 20. You would want to ask yourself if you were able to go beyond the current setting range on Setting A, (about 5-25) to further minimize the Output. In some cases, this will not be possible for considerations outside of the experiment, e.g., safety, cost, productivity, etc. You will find in most mature industries that if there is a problem that cannot be solved it is likely due to an interaction. It is easy to pick up a direct effect, such as Output that goes down as you turn up Setting B, so you will likely already know these effects. But the human brain is not set up to recognize even a two-way interaction (as above) much less more. The only way to catch these effects and discriminate them from random effects is through statistical analyses.