In multifactorial analyses, also known as the method of multiple working hypotheses, we acknowledge that many different processes are operating to shape ecological patterns. Our goals are to determine which effects can be distinguished from noise; evaluate the relative strength of different patterns; and understand their interactions. Information-theoretic (IT) and multi-model averaging (MMA) approaches are widely used but suboptimal tools for pursuing a multifactorial approach in ecology. Conceptually, IT approaches encourage ecologists to test a series of straw-man hypotheses by inventing a series of discrete sub-models where a subset of the processes are assumed to be absent. It is well-known that simple post-selection inference - i.e., naively computing p-values or confidence intervals based on a single model chosen by IT (or other) methods - is misleading. Evaluating variable importance by comparing summed IT weights has little grounding in theory, and is harder to interpret than simply comparing the scaled parameter estimates and confidence intervals from the full model. MMA improves on simpler IT approaches by implementing a simple form of shrinkage estimation (a way to improve parameter estimation and prediction by "shrinking" small parameter estimates toward zero). An unappreciated disadvantage of shrinkage methods is that post-shrinkage confidence intervals (CIs) are at best difficult to derive and at worst no better than the CIs obtained from the full model. In addition, newer approaches to shrinkage estimation are faster and have clearer statistical underpinnings than MMA. If researchers want accurate estimates of the strength of multiple competing ecological processes along with reliable confidence intervals, their best hope is to use full (maximal) statistical models after making principled, a priori decisions about which predictors to include.