The reproducibility of frequency measurements using TremAn was comparable to the accelerometry, with the TremAn/accelerometry ratio of measurement error standard deviations equal to 0.99 (95% confidence interval (0.84, 1.17)).
The TALA implementation introduced herein implicitly assumes that the measurement error standard deviations are known.
In cases for which field testing is not practicable, or in extended operating conditions, measurement error standard deviations may not be accurately known.
To illustrate the varying measurement error standard deviations used in the simulation scenarios, we show in Figure 1 observed longitudinal measurements from the same 100 patients with σ e = {0.1,0.5,1}, and when α = 0.25. Figure 1 illustrates that as the measurement error standard deviation increases, the variability in the observed biomarker values increases.
where (sigma _{tau } triangleq sqrt {2}csigma _{t}) is the DDOA measurement error standard deviation; and γ≥0 is a pre-specified multiplier.
Supplementary Fig. 1 shows the power of the four end-points when the measurement error standard deviation is changed.
The measurement error standard deviation (see supplementary methods for more details) is varied between 0 and 0.25, corresponding to no and extremely high measurement error respectively.
The measurement error standard deviation s and the system noise factor σ are identical to the true values using the SDE approach.
Hence, we conclude that when the system noise is set to zero (equivalent to the ODE case), the measurement error standard deviation is increased to account for that variability, which is not unexpected since we neglect the variability in dynamics.
Under the standard Weibull model, we observe increasing bias in the estimates of the association between baseline biomarker values and survival, as the magnitude of the measurement error standard deviation, σ e, increases.
