Microscopic modeling

We develop mechanistic models to understand and predict the performance of the different technologies under development at the EERL. We combine the models with parameter estimation techniques and modeling discrimination techniques. Microscopic Modeling software and hardware include:

1. Visual PEST

2. DATA fit

DataFit is a science and engineering tool that simplifies the tasks of data plotting, regression analysis (curve fitting) and statistical analysis. The software presents the following features:

• Multivariate linear or nonlinear regression. DataFit can solve linear and nonlinear regression models with up to 20 independent variables.

• Variable Selection (Data Mining). DataFit includes Forward Selection, Backward Elimination, Stepwise Selection and Manual variable selection modes to help determine which independent variables should be included in your regression model.

• Pre-defined regression models – there are currently 60 two-dimensional and 242 three-dimensional nonlinear regression models pre-defined in DataFit.

• User defined regression models

• Robust solver – DataFit utilizes the Levenberg-Marquardt method with double precision to perform nonlinear regression.

• Various solution options – You can perform linear OR nonlinear regression on one model at a time that you choose from a list, groups of models (pre-defined model groups, or define your own), or all of the models available with one mouse click.

• Automatic solution ranking - As regression models are solved, they are sorted automatically according to the goodness of fit criteria you specify (Residual Sum of Squares, Correlation Coefficient, DOF Adjusted Correlation Coefficient or Standard Error). If you aren't an expert and aren't sure which equation you should use to model your data, DataFit helps you make this decision.

• Detailed regression results – DataFit reports the following information about each solved regression model automatically without additional user action:

  - Number of observations

  - Number of missing observations

  - Number of nonlinear iterations performed

  - Sum of Residuals

  - Average Residual

  - Residual Sum of Squares

  - Standard Error of the Estimate

  - Coefficient of Multiple Determination (R2)

  - Proportion of Variance Explained

  - Adjusted coefficient of multiple determination (Ra2)

  - Durbin-Watson statistic

  - Coefficient values, Standard Error, t-ratio and Prob(t)

  - Confidence Intervals (68, 90, 95 and 99 percent levels)

  - Variance Analysis (ANOVA)

  - Robust computers (6 computers)