|The GoM Concept
GoM methods have been applied successfully to human populations where multiple measures of health, prevalence of various risk factors, and morbidity or mortality outcomes are observed on each individual. The GoM model has been applied to studies in many fields; studies include prevalence of genetic markers, sensory perception, analysis of psychiatric conditions, establishment of shadow prices, among others. It also has wide ranging business applications, including exploratory data analysis, data mining, and market segmentation.
GoM has been used for the analysis in a large number of research projects primarily at Duke University’s Center for Demographic Studies. Decision Systems Inc., under the leadership of Dr. Gene Lowrimore is developing a software product, DSIGOM to make the capabilities of GoM more widely available.
At decision systems, inc., we welcome any feedback you have regarding DSIGoM. Please email your thoughts to firstname.lastname@example.org.
|Features of DSIGOM Beta Version 1.01|
- Facilities are provided for describing the GoM variables
One can define sets of outcomes as named objects that can be assigned to a set of variables with a couple of mouse clicks.
Before invoking the GoM engine an extensive verification process runs
- short and long name
- internal, external, indicator, or identity
- whether in or out of the current analysis
The system includes facilities for both flavors of GoM regression, i.e.
- the specified number of pure types is verified against the cardinality of the outcome set attached to the indicator variable,
- every included variable must be designated as internal or external and have an outcome set attached,
- the specified format of the file is verified against the record length of the file.
The computational engine is an implementation of the p-GoM model and is essentially the same as that which is the basis of most of the GoM analyses published by the Center for Demographic Studies.
The software includes a Help System and a User Guide, both of which are in hypertext format and available from the DSIGoM main menu.
Outputs consist of a standard report, a file of the gik values and a file of the probability profiles. These latter two are provided to permit further analysis and report preparation in SAS, Excel, Word, etc.
- solving for the probabilities conditional on the grade of membership structure (giks) being held fixed, and
- solving for giks conditional on the probability profile structure being fixed.