Glossary of Terms  
 
 

Receiver Operating Characteristics (ROC) used to describe CHEMSAS™ results: 95% Confidence Interval: The numeric range between which model predictions would be expected to fall 95% of the time. In other words model predictions would be expected to fall OUTSIDE of the 95 % Confidence Interval only 1 in 20 times. Accuracy: Total number correct out of total sample size. Area under the Receiver Operating Curve (ROC): This is a statistical measure of how well the model distinguishes between possible outcomes. Values vary between 0.0 and 1.0 with an Area under ROC = 1.0 meaning that the model can perfectly distinguish between possible outcomes. Insilico: Performed as a digital algorithm. Compare to Invitro which literally translates into Performed in glass, referring to the traditional testtube based assays. Negative Predictive Value (NPV): The accuracy of predictions that the target outcome is NOT PRESENT. Positive Predictive Value (PPV): The accuracy of predictions that the target outcome is PRESENT. R Squared: This is the Coefficient of Multiple Determination and is a statistical estimate of how well a model fits actual data. In simplest terms the R squared value estimates how much of the variability in the observed data can be explained by the model. For example, if model predictions had an R squared of 0.90 then approximately 90% of the variability in the data can be predicted from the model. Or in other words about 10% of the variability in the data is not explained by the model. Note: This is not the same as the square of the linear correlation coefficient! Sensitivity: The percentage of cases predicted to have the target outcome when the outcome is PRESENT  eg. positive in disease. Specificity: The percentage of cases predicted not to have the target outcome when the outcome is NOT PRESENT  eg. negative in health.

Critical Outcome Technologies Inc., © 20002001. All rights reserved. 