|
|
Specifications
Model Fitting |
| Nonparametric Multiplicative Regression |
- Local mean - uniform weights in window
- Local mean - Gaussian weights
- Local linear - Gaussian weights
|
| Least-squares Multiple Linear Regression |
- Direct entry
- Backwards elimination
- Cross-validation
- Monte-carlo test
|
| Overfitting and Parsimony Controls |
- Built-in leave-one-out cross validation
- Minimum average neighborhood size
- Data:predictor ratio
- Improvement criterion
- Monte Carlo test
|
Prediction |
- GIS application - input of predictors grids, output of response grids
- Predict responses for new cases
- Estimate responses to categorical variables
|
Data Types |
| Response Variables |
- presence/absence
- quantitative
|
| Predictor Variables |
|
|
Data Modifications |
- Power transform
- Log
- Arcsine squareroot
- Delete rows or columns
- Multiply or add a constant
- Shuffle within columns
- Matrix editing
|
Data Formats |
- *.wk1
(easy import/export from Excel and most statistical software)
- File structure compatible with PC-ORD
- Import/export *.xls, *.xlsx, *.csv format
- GIS grid file (*.asc) input/output
|
GIS Functionality |
- Read multiple input grids
- Write response grids for import to ESRI software
|
Graphs |
- Scatterplots of observations, estimates, and residuals
- Simple scatterplots among all pairs of variables
- Matrix scatterplots
- 2D response curves
- 3D response surfaces
- 3D contour plots
- 3D rotation
- Save as Windows metafile or paste into Office document
- Color or black/white
|
Help System |
- Context sensitive
- Detailed and extensive
- Companion pdf document explaining the rationale and detailing the underlying
math
|
Limitations |
- 30 GIS input files (one file for each predictor variable)
- 100 GIS output files (one file for each species)
- 246-character file names
- 1000 species
- 500 habitat variables in the model at once (more can be included as potential
variables)
- 32,000 sample units (rows)
|
System Requirements |
- Operating System: Windows 95, 98, NT, ME, 2000, XP
- 80486 or higher CPU (including Pentium 4, Athlon, Celeron, etc.)
- 8 MB RAM (more RAM means ability to analyze larger data sets)
- 6 MB of available hard disk space
|
|
