Leaderboard

MLR uses happenstance data gathered, in this case, there is no guarantee historical data contains all the factors in which we are interested. We might miss some factors. It is also possible we want to check to look at three variable interactions but data have only two-level interactions. Happenstance data may contain some noise factors. if happenstance data contain a noise factor it can mislead the interaction factor, we cannot get better information about the interaction effect. DOE – If we are creating experimental data we can control the variable and check the interaction between variable factors by controlling all noise factors. This will give us a correct and authentic interaction effect. By choosing DOE will give a clear-cut interaction effect. With DOE you can do blocking & noise treatments to ensure signals comes from the factors. With DOE we would have more control & accurate measurement system compared to MLR.