Prediction and Cross-Validation of Video Cartographic Symbol Location Performance

Abstract

A response surface methodology central-composite design was used to obtain multiple-regression prediction equations of performance on a video cartographic symbol search task. Observers were required to locate the position of designated target symbols on a series of maps displayed on black-and-white and color television (TV) monitors. The variables used to predict both location and latency performance were focus, density of nontarget symbols, visual angle of the observer, and TV raster lines per mm of actual map area. Prediction equations were compared for black-and-white and color TV monitors through collapsed and uncollapsed, within-subject data analyses. Both analysis procedures were compared in terms of resulting sensitivity and in terms of the predictive validity of the regression equations as determined in cross-validation. It was concluded that the uncollapsed, within-subject designs provided the better prediction equations.