Statistical analysis is valuable because it is used to identify trends in the original data, but too much data manipulation is referred to as overfitting. There is an optimal level of analysis. Too little and trends are missed. Too much and trends are artificially manufactured. The most appropriate statistical tools are powerful in part because they are directly related to the original data and do not require excessive manipulation. Machine learning or artificial intelligence use algorithms to find trends. This bioinformatic approach is currently popular, and it has exceptional usefulness when time is an issue or when the dataset is large. But the inherent assumptions and the processes are opaque and can arrive at inappropriate or irreproducible conclusions. Complex statistical modeling by multivariate correlational analysis such as principal component analysis or clustering techniques are also popular, but they are also inaccessible to most biomedical investigators. Consequently, there is a tendency to farm out data analysis. Yet, this step is crucial for understanding the meaning in the data and making sound interpretations. Two prior ideas are synthesized at this point: the faith-based dependency of artificial intelligence and predictive modeling and the need for explicit validation to conform to accepted scientific standards. Data analysis is explained without equations so that it is less intimidating. The impetus of this essay is to provide confidence to biomedical investigators, to ameliorate the investigator’s disconnect with the original data, and thereby to prompt the biomedical scientist to become the auteur for the investigation.