Gs. Martin et al., VALIDATION OF A REGRESSION-MODEL FOR STANDARDIZING LIFETIME, Journal of the American Veterinary Medical Association, 210(11), 1997, pp. 1641
Objective--To determine the relationship between prediction errors of
a regression model of racing finish times and earnings or finish posit
ion; the relationship between standardized finish times, determined by
use of this model, and earnings or finish position; and whether this
model was valid when applied to data for horses that underwent surgica
l treatment. Design--Survey. Sample Population--Records of 6,700 healt
hy Thoroughbreds racing in Louisiana and of 31 Thoroughbreds with idio
pathic left laryngeal hemiplegia that underwent surgical treatment. Pr
ocedure--Predicted and standardized finish times were calculated by us
e of the regression model for healthy horses, and the relationships be
tween prediction error (actual - predicted finish time) and standardiz
ed finish times, and earnings and finish position, were examined. Then
, the regression model was applied to data for horses with hemiplegia
to determine whether the model was valid when used to calculate predic
ted and standardized finish times for lifetime performance data. Resul
ts--Prediction error and standardized finish times were negatively cor
related with earnings and positively correlated with finish position a
nd, thus, appeared to be reliable measures of racing perfor mance. The
regression model was found to be valid when applied to lifetime perfo
rmance records of horses with laryngeal hemiplegia. Clinical Implicati
ons--Prediction error and standardized finish times are measures of ra
cing performance that can be used to compare performances among Thorou
ghbred racehorses across a variety of circumstances that would otherwi
se confound comparison.