Rice Analytics

Automated Reduced Error Predictive Analytics




Consumer Survey with Many Variables
With small sample sizes and the large number of correlated variables often observed in survey research, Reduced Error Logistic Regression (RELR) software allows accurate models with fewer worries about "overfitting" and multicollinearity problems. In consumer surveys involving branding, positioning or customer satisfaction/loyalty variables, a limiting factor is often that the analytic approach does not support the number of variables in the survey.  The result is that the conclusions of the research can be compromised.  Please visit the Case Studies page of this website to see how RELR overcomes these problems with real survey data in several case studies that have been presented at conferences.



A model built from a small sample size with Reduced Error Logistic Regression can be more accurate and much more interpretable than a model built with a sample size that is 10-100 times larger with standard predictive modeling approaches.  Obviously, this can save you money in data collection costs. 

Machine Learning  Segmentation  Consumer Surveys  Predictive Modeling  Risk Management