Revamping Data Analytics #22
Description
Main idea: Create another chapter on the websites for this class or CBE 20258/60258 to cover the following topics.
- Maximum likelihood estimation (MLE) as a lens to explain why weighted nonlinear regression works
- Fisher information matrix (FIM) derived from the MLE perspective with simplifications for i.i.d. Gaussian measurement errors
- Parameter covariance estimate derived from MLE and optimization perspectives. The goal is to explain in two ways when the formulas from CBE 20258 work.
- MLE for (1) proportional plus constant or (2) auto-correlated measurement errors
- Eigendecomposition of FIM for practical identifiability analysis
- Model-based design of experiments using scipy
- Link to ParmEst and Pyomo.DoE tutorials from the summer workshop