Table 2 Mean Brier Score and mean scaled Brier Score (%) for the assessment of candidate binomial GAMMs. Model selection was conducted by removing non-significant variables with backward selection. Models were assessed using 5-fold cross validation. Standard deviation is given in brackets. The removal of the variables sex and race type result in minor changes in the performance of the models as given by the Brier Score. Hence their inclusion in the model was not supported because they did not improve model performance. EHI refers to exertional heat illness.

From: Risk factors for, and prediction of, exertional heat illness in Thoroughbred racehorses at British racecourses

Model

Brier Score

Scaled Brier Score

\(\Delta\) Scaled Brier

M1: Full\(^{\rm a}\)

0.0148 (0.0007)

16.9446 (0.2706)

 

M2: M1 - Sex\(^{\rm b}\)

0.0149 (0.0006)

17.0211 (0.2906)

0.0765

M3: M2 - Race Type\(^{\rm c}\)

0.0149 (0.0006)

17.0019 (0.2813)

-0.0192

  1. \({}^{\rm a}\) \(EHI \sim s(Age) + s(Distance) + s(WBGT) + s(5DayAverage) + Going + PreviousInc + Year + RaceType + RaceOff + Sex\)
  2. \({}^{\rm b}\) \(EHI \sim s(Age) + s(Distance) + s(WBGT) + s(5DayAverage) + Going + PreviousInc + Year + RaceType + RaceOff\)
  3. \({}^{\rm c}\) \(EHI \sim s(Age) + s(Distance) + s(WBGT) + s(5DayAverage) + Going + PreviousInc + Year + RaceOff\).