Fig 1.
Normally distributed data show an oscillating pattern when plotted over time.
Green (light): mean = 0, SD = 150. Blue (dark): mean = 0, SD = 350. Density plots (left hand panel) based on 10 000 randomly generated data points with a standard normal distribution (using Rnorm()). The lines (right hand panel) show the first 30 data points of each dataset.
Fig 2.
Observed average values of BI, ABV, TL-BS variability, and Average TL-BS positive for each study and each run of the first three simulation series (SD, mean, and bias increasing).
Each data point represents the group average for a single study, lines represents the average observed for each run of 1000 studies. The shades of the data points (but not the lines) indicate the percentage significant group differences observed per run (1000 studies).
Fig 3.
TL-BS time series and traditional BI for increased SD, mean, and bias.
Panels B, C, and D show data for the 'change group' in the 1000th study of the 10th run in the increasing SD, mean, and BI simulations respectively. Similar to Fig 2 reported by Zvielli and colleauges [1], a smoothing procedure was applied to the TL-BS data.
Fig 4.
Observed average values of BI, ABV, TL-BS variability, and Average TL-BS positive for each study and each run of the dynamic bias simulations.
Each data point represents the group average for a single study, lines represents the average observed for each run of 1000 studies. The shades of the data points (but not the lines) indicate the percentage significant group differences observed per run (1000 studies).
Fig 5.
ABV time series (BI per bin) and traditional BI when increasing dynamic bias switching frequency and increasing dynamic bias magnitude.
Each row depicts data for the ‘change group’ in the 1000th study of runs 1, 5, and 10 of the increasing dynamic frequency simulation (top) and the increasing dynamic magnitude simulation (bottom).