Search

The intent of this study was to understand via simulation how data from designed experimentation for linear models can succeed in the prediction of individual values despite its relatively small size which renders data splitting for validation purposes nonviable. Another intent was to emphasize why, for a given level of precision, designed experimentation requires far many more runs for the prediction of individual values than it does for its more mundane use of mean prediction, and how this required number of runs can be determined via simulation as a function of the model validation method used. The results showed that prediction with designed data can be successful given its low tendency to overfitting and that model reduction can be detrimental to prediction which contrasts with the pursuit of a bias-variance tradeoff with undesigned data. As designed data increasingly resembles undesigned data, either by containing factors that have zero influence in the response, having high correlation among factors levels, and/or having small n to p ratio, model reduction becomes increasingly necessary for prediction. [ABSTRACT FROM AUTHOR]

The large availability of undesigned data, a by-product of chemical industrial research and manufacturing, makes it attractive the venturesome use of machine learning for its plug-and-play appeal in attempt to extract value out of this data. Often this type of data does not only reflect the response to controlled variation but also to that caused by random effects. Thus, machine learning based models in this industry may easily miss active random effects out. This study shows by simulation the effect of missing a random effect via machine learning — vs. including it properly via mixed models as a benchmark — in a context commonly encountered in the chemical industry — mixture experiments with process variables — and as a function of relative cluster size, total variance, proportion of variance attributed to the random effect, and data size. Simulation was employed for it allows the comparison — missing vs. not missing random effects — to be made clear and in a simple manner while avoiding unwanted confounders found in real world data. Besides the long-established fact that machine learning performs better the larger the size of the data, it was also observed that data lacking due specificity — i.e. without clustering information — causes critical prediction biases regardless the data size. [ABSTRACT FROM AUTHOR]