Information Graphs Incorporating Predictive Values of Disease Forecasts.

Area of Science:

• Plant pathology
• Forecasting science
• Decision analysis

Background:

• Diagrammatic formats are crucial for evaluating forecasts in plant pathology and disease management.
• Decisions on interventions often rely on proxy risk variables from forecasts.
• Existing formats have unexploited information properties.

Purpose of the Study:

• Introduce a novel diagrammatic format for disease forecasts.
• Characterize forecasts using relative entropies and predictive values.
• Complement existing forecast evaluation methodologies like ROC curves.

Main Methods:

• Developed a new diagrammatic format with two categories of actual status and two of forecast.
• Calculated relative entropies, functions of predictive values.
• Utilized data requirements identical to those for Receiver Operating Characteristic (ROC) curve calculation.

Main Results:

• The new format displays relative entropies, quantifying expected information from forecasts.
• It characterizes forecasts based on predictive values, analogous to ROC curves using sensitivity and specificity.
• The method requires no additional data beyond ROC curve calculations.

Conclusions:

• The proposed diagrammatic format offers a complementary approach to ROC methodology for forecast evaluation.
• It provides insights into forecast performance based on relative entropies and predictive values.
• This enhances the comparison and evaluation of disease forecasts for management decisions.