Skip to main content

Table 4 Binary logistic regression models and their evaluations

From: A new postural stability-indicator to predict the level of fear of falling in Parkinson’s disease patients

(a) Binary logistic regression outcomes based on FTR1,FTR2, and FTR1/2 parameters
Models Odds ratio p-value 95% CI for odds ratio
Lower Upper
 FTR1 1.54 0.005* 1.14 2.07
 FTR2 1.65 0.01* 1.13 2.42
 FTR1/2-ground 33.55 0.005* 2.91 386.72
 FTR1/2-40 cm 30.55 0.005* 2.78 335.73
(b) Receiver operating characteristic (ROC) curve analysis and corrected Akaike’s Information Criterion (CAIC) of FTR1, FTR2, and FTR1/2
  FTR1 FTR2 FTR1/2-ground FTR1/2-40 cm
 AUC 0.921 0.932 0.939 0.921
 95% CI for AUC 0.830 to 1 0.845 to 1 0.860 to 1 0.835 to 1
 p-value < 0.0001** < 0.0001** < 0.0001** < 0.0001**
 Overall accuracy 86.84 89.47 92.1 89.5
 Sensitivity 89.29 92.9 96.43 96.43
 Specificity 90 80 80 70
 CAIC 24.1 22.58 21.56 22.9
Optimal cutoff value ≤ 59.43 > 21.36 ≤ 2.83 ≤ 2.71
(c) Cross-validation results
 Error 15.79 21.05 10.53 13.16
 Sensitivity 82.14 71.43 92.86 92.86
 Specificity 90 88.57 80 70
  1. AUC Area Under the Curve, CI Confidence Interval, FTR Functional Time Ratio
  2. **p-value ≤ .0005; *p-value < .05