Three common statistical missteps we make in reservoir characterization

Reservoir characterization analysis resulting from incorrect applications of statistics can be found in the literature, particularly where integration various disciplines is needed. Here, we look at three misapplications ordinary least squares linear regression (LSLR) and show how they lead to poor results offer better alternatives, available. The issues are 1. Application algebra an LSLR-derived model reverse roles explanatory response variables that may give biased predictions. In particular, examine pore throat size equations (e.g., Winland’s Pittman’s equations) find claims over-predicted permeability part due statistical mistakes. 2. Using a log-transformed variable LSLR model, de-transforming without accounting for role noise gives equation which under-predicts mean value. Several approaches exist address this problem. 3. Mis-application R2 cases misleading results. For example, fitting decline curve optimistic values, as also case multimodal present. Using actual synthetic datasets, illustrate effects these errors have on some implications using machine learning