Table 3 Stepwise backward selection regression model to predict medication state transitions, ordered by relative importance

From: Digital speech biomarkers can measure acute effects of levodopa in Parkinson’s disease

 

Estimate

p value

Medication change Model (dependent variable: Change OFF to ON or ON to OFF)

INTERCEPT

−0.34041

0.51172

LTAS mean (text) change

0.04163

<0.01 **

MPT (phon) change

0.66034

<0.01 **

Int kurt (text) change

−3.39710

<0.05 *

RST (text) change

−0.02887

0.05147

LTAS skew (text) change

0.54767

0.14261

sdF0 (text) change

2.54548

0.14421

NSR (text) change

1.48410

0.2086

Null deviance: 70.681 (50 df); Residual deviance: 29.23; AIC: 45.23

  1. NSR Net speech rate, RST Rate of speech timing, Int Kur Kurtosis of Intensity, sd F0 standard deviation of fundamental frequency, MPT Maximal phonation time, LTAS Long-term averaged spectrum, sd standard deviation, kurt kurtosis, skew skewness.
  2. p-value significance—*p < 0.05, **p < 0.01.
Back to article page