SIMFIT - A Computer Package for Simulation, Curve Fitting and Statistical Analysis Using Life Science Models

Computers are widely used for sensitivity analysis, exploring the behavior of model systems, statistical methods of data processing and curve-fitting. However, all these processes are controversial. Simulation is of limited value unless realistic density functions are employed and sufficient iterations are used to achieve stable results. Statistical analysis is complicated by using methods which are only justified with linear, normal models and constant variance, so their robustness and bias has to be estimated independently. Curve-fitting is complicated when there are insufficient replicates for accurate weighting and smoothed weighting schemes have to be used, or when parameter constraints have to be satisfied during the optimization. Finding starting estimates, internal parameter and data scaling, avoiding local minima of the objective function and methods to be used for differential equation solving or optimization are also problematical with constrained nonlinear weighted least squares regression. Also parameter redundancy is frequently encountered.