MobyDigs
MobyDigs is a software package that has been developed by the Milano Chemometrics and QSAR Research Group for the calculation of regression models by using genetic algorithms for variable selection to obtian an optimal subset of predictive models. The problem in fact is: how to develop reliable regression models when you have 100, 500, 1000, 2000 candidate variables?
MobyDigs - version 1.1 - is now commercially available and is released in the version MobyDigs Professional, running on Windows platforms.
What you can do by MobyDigs?
# Calculate a single regression model by Ordinary Least Squares regression (OLS).
# Obtain a set of optimal regression models by OLS.
# Perform further validation of the final regression models
# Add new objects and/or new variables to the previous data
# Perform predictions of external data by using the obtained models
# Get a final subset of modelling variables and use them for other purposes
# Develop Principal Component Regression (PCR) models
# Use extended diagnostic and graphic tools for analysing each regression model
# Evaluate similarity/diversity among the final obtained models
# Evaluate average predictions from 2 or more final models
# Variable analysis by powerful graphic tools and statistic parameters
# Save any kind of results in tabbed text files
# Build - automatically - a final report of your work, graphics included
Read an example of application.
MobyDigs characteristics
# Import of any kind of ASCII data structure formats
# Maximum number of samples: user defined (*)
# Maximum number of variables: user defined (*)
# Maximum number of populations: 10
# Maximum number of final retained models: 100
# 7 different regression parameters for model optimization (Q2, R2adj, LOF, ...)
# On-line managment of the whole GA evolution process
# Final validation procedures: boostrap, Y-scrambilng, external data
# Preliminary exclusion of probably not modelling variables by a Tabu list
# Recovery of the variables previously put into the Tabu list
# Consensus analysis, i.e. predictions performed as averages from different models
# Evaluation of the similarity/diversity of the final models
# Single regression model without any GA development
# Different views of variables and models, statistics, professional graphics
# A comprehensive user manual including technical definitions is available
(*) The maximum allowed number of objects and variables is not fixed, but can be tuned by the user. The available options are:
1. Object / Variables: 2500 / 2500
2. Object / Variables: 2000 / 4000
3. Object / Variables: 1000 / 5000
4. Object / Variables: 10000 / 1000
In the MobyDigs Evaluation version all the options and tools are fully working. The unique restriction concerns tha data sets: in the evaluation version, only some predefined data sets can be used. These are well known published data sets, allowing to apply any kind of modelling strategy and check results. Bibliographic references and characteristics of the proposed data sets are also provided. Download the evaluation version of MobyDigs.
What's new in MobyDigs 1.1
Look the advanced MobyDigs features, the system requirements and the theoretical background.
MobyDigs is a software package that has been developed by the Milano Chemometrics and QSAR Research Group for the calculation of regression models by using genetic algorithms for variable selection to obtian an optimal subset of predictive models. The problem in fact is: how to develop reliable regression models when you have 100, 500, 1000, 2000 candidate variables? MobyDigs - version 1.1 - is now commercially available and is released in the version MobyDigs Professional, running on Windows platforms. What you can do by MobyDigs?# Calculate a single regression model by Ordinary Least Squares regression (OLS).
# Obtain a set of optimal regression models by OLS.
# Perform further validation of the final regression models
# Add new objects and/or new variables to the previous data
# Perform predictions of external data by using the obtained models
# Get a final subset of modelling variables and use them for other purposes
# Develop Principal Component Regression (PCR) models
# Use extended diagnostic and graphic tools for analysing each regression model
# Evaluate similarity/diversity among the final obtained models
# Evaluate average predictions from 2 or more final models
# Variable analysis by powerful graphic tools and statistic parameters
# Save any kind of results in tabbed text files
# Build - automatically - a final report of your work, graphics included
Read an example of application. MobyDigs characteristics# Import of any kind of ASCII data structure formats
# Maximum number of samples: user defined (*)
# Maximum number of variables: user defined (*)
# Maximum number of populations: 10
# Maximum number of final retained models: 100
# 7 different regression parameters for model optimization (Q2, R2adj, LOF, ...)
# On-line managment of the whole GA evolution process
# Final validation procedures: boostrap, Y-scrambilng, external data
# Preliminary exclusion of probably not modelling variables by a Tabu list
# Recovery of the variables previously put into the Tabu list
# Consensus analysis, i.e. predictions performed as averages from different models
# Evaluation of the similarity/diversity of the final models
# Single regression model without any GA development
# Different views of variables and models, statistics, professional graphics
# A comprehensive user manual including technical definitions is available
(*) The maximum allowed number of objects and variables is not fixed, but can be tuned by the user. The available options are:
1. Object / Variables: 2500 / 2500
2. Object / Variables: 2000 / 4000
3. Object / Variables: 1000 / 5000
4. Object / Variables: 10000 / 1000
In the MobyDigs Evaluation version all the options and tools are fully working. The unique restriction concerns tha data sets: in the evaluation version, only some predefined data sets can be used. These are well known published data sets, allowing to apply any kind of modelling strategy and check results. Bibliographic references and characteristics of the proposed data sets are also provided. Download the evaluation version of MobyDigs. What's new in MobyDigs 1.1 Look the advanced MobyDigs features, the system requirements and the theoretical background.