Summary statistics for both combined fall risk models as the number of steps used changes.

indicates the model that excludes the long-term Lyapunov exponents, and indicates the model that includes this data. Each column indicates the number of steps used in the fall risk model. For each number of steps n, a simulation was created (usually taken from a longer simulation), fall risk metrics were calculated from this simulation, the PCA transformation was applied, and then ln(MFPT) was estimated using Eq 12. The values are over all of the trials of a given number of steps n. The root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), relative average absolute error (RAAE), and maximum absolute error (Max Abs Err) are given. The % in Tolerance indicates the percentage of these trials in which the predicted value was within the true ln(MFPT)±3.3. The mean SD are the mean of the standard deviation data for each training condition. The slope, intercept, and are each from a linear fit of the true and estimated ln(MFPT) for all trials (Eq 15). Values of 1, 0, and 1, respectively, would indicate a perfect fit. Cells with no value indicate that there was not enough data to calculate these statistics for the given number of steps.