Morrisscreening

Morrisscreening is a rapid variable screening approach used in mathematical modelling and data analysis to identify input factors that most influence a model’s output. The term is commonly used to describe a family of screening techniques that build on the Morris method for global sensitivity analysis and may appear in modelling workflows as a preliminary step to reduce dimension before more detailed analysis.

Conceptually, Morrisscreening involves defining input ranges for a set of factors, generating a series of quasi-random trajectories through the input space, evaluating the model along each trajectory, and extracting information about how individual inputs affect the output. The analysis typically reports elementary effects, summarized by statistics such as the mean of absolute elementary effects (μ*) and the standard deviation (σ). Factors with high μ* are considered influential; high σ suggests nonlinear responses or interactions with other factors. In practice, a pre-screening cutoff or ranking is often used to narrow a large set of inputs to a manageable subset for subsequent, more exhaustive analysis.

Applications of Morrisscreening are common in engineering, environmental modelling, pharmacokinetics, and systems biology, where models can be expensive to evaluate and the number of input variables is large. The method is valued for its computational efficiency and its ability to detect nonlinearities and potential interactions in a relatively small number of model runs.

Limitations include its qualitative nature rather than precise quantitative estimates of sensitivity, dependence on sampling design and chosen levels, and potential to overlook factors with subtle effects or those affected by significant noise. Morrisscreening is typically used as a first screening step, followed by more rigorous sensitivity analyses on the selected subset.

See also Morris method; global sensitivity analysis; screening methods.