Skip to main content

Table 5 Effect of sampling rate on model performance

From: Role of data measurement characteristics in the accurate detection of Parkinson’s disease symptoms using wearable sensors

Sampling Rate (Hz)

Tremor

Bradykinesia

Binary

Multiclass

Binary

Multiclass

62.5

0.77 (0.67, 0.87)

0.74 (0.65, 0.82)

0.67 (0.61, 0.74)

0.65 (0.59, 0.71)

50

0.77 (0.67, 0.87)

0.74 (0.66, 0.83)

0.68 (0.61, 0.74)

0.65 (0.59, 0.70)

40

0.77 (0.67, 0.87)

0.75 (0.66, 0.84)

0.68 (0.61, 0.74)

0.65 (0.59, 0.70)

30

0.76 (0.66, 0.86)

0.74 (0.66, 0.83)

0.68 (0.61, 0.74)

0.65 (0.59, 0.70)

20

0.75 (0.65, 0.85)*

0.73 (0.64, 0.81)

0.67 (0.61, 0.74)

0.65 (0.59, 0.70)

10

0.73 (0.64, 0.82)*

0.70 (0.62, 0.78)*

0.68 (0.61, 0.74)

0.64 (0.58, 0.70)

7.5

0.72 (0.63, 0.81)*

0.69 (0.61, 0.77)*

0.69 (0.62, 0.75)

0.65 (0.60, 0.70)

5

0.70 (0.62, 0.79)*

0.70 (0.62, 0.78)*

0.67 (0.60, 0.74)

0.65 (0.60, 0.71)

  1. Average and 95% confidence intervals of model performance (AUROC) to classify PD symptoms using different sampling rates for the Accel (tremor) or Combo (bradykinesia) sensor types. Asterisk (*) indicates significant difference from performance at the original sampling rate. Bolded results indicate the sampling rate selected for subsequent analyses