LOTUS version 2.3    

Logistic Regression Tree with Unbiased Selection

 

LOTUS is a logistic regression tree algorithm developed by Kin-Yee Chan and Wei-Yin Loh (University of Wisconsin-Madison).

LOTUS is unique among logistic regression tree algorithms in possessing the following features:

Negligible bias in variable selection (very important for tree interpretation);
Ability to use ordered (continuous) and unordered (categorical) predictor variables;
Choice of roles for predictor variables (splitting only, node modeling only, both, or none);
Choice of piecewise best simple linear, multiple linear or stepwise logistic regression models;
Choice of stopping rules: pruning by cross-validation or prunning with a test sample;
Automatic handling of missing values;
Automatic generation of LaTeX (MikTeX) or allCLEAR source code for the tree diagrams. The LaTeX code requires the PSTricks package.

Documentation:

  1. Chan, K.-Y. and Loh, W.-Y. (2004), "LOTUS: An algorithm for building accurate and comprehensible logistic regression trees," Journal of Computational and Graphical Statistics, 13(4): 826-852. [This is the principal reference for LOTUS] [postscript] [pdf]
  2. Loh, W.-Y. (2006), " Logistic regression tree analysis," Handbook of Engineering Statistics, H. Pham, Ed. Springer, 537-549. [pdf]
  3. LOTUS User Manual in postscript or pdf format. The manual uses the example data and description files car.dat and car.dsc for illustration.

Compiled binaries: The following files are freely distributed for non-profit use only.

Intel and compatibles (Windows 9x/NT/2000/XP) in winzip format --- download
Intel and compatibles (Linux 2.4 or later) in gzip format --- download

Revision history: See the file history.txt

Related tree algorithms with unbiased selection:

GUIDE: Classification and regression tree
CRUISE: Classification trees with multiple splits at each node
QUEST: Binary classification tree

File viewers:

Ghostscript and GSView: view postscript files
Adobe Reader: view Adobe portable document format (pdf) files
Last updated: January 27, 2024