- This event has passed.
Ben Randall, NC State, Introduction to local and global sensitivity analysis
September 20, 2018 | 3:00 pm - 4:00 pm EDT
Mathematical modeling is a way for researchers to visualize components of a system and analyze how those components interact. When these models become significantly large, many problems arise. Here are a few:
1. As models increasingly complex with a myriad of parts, it is prudent for the researcher to ask whether all of these parts are necessary to describe the system of interest. Could you have described your system, eliciting your quantity of interest, with a smaller model without sacrificing (many) model assumptions?
2. When the model is optimized to fit data for a system, not every parameter can be informed by the data available. Typically, the researcher must perform a subset selection to determine a set of parameters to estimate.
Sensitivity analysis can aid in the identification of model parts and parameters that are insensitive to the researcher’s quantity of interest, which can facilitate model simplification and parameter estimation. This talk will develop what are known as “global” sensitivity methods and apply them to a neurological model describing the Valsalva maneuver.