Table 1.
Parameters used in the simulations.
Values in italics are the baseline scenario.
Fig 1.
False positive probabilities by time (research cycle) for four levels of cycles per audit (j).
The grey lines are 20 randomly selected simulations and the dark red line is the average of 500 simulations.
Fig 2.
Mean number of papers produced (left) and mean effort (right) over time (research cycle) for 500 simulated labs for four levels of cycles per audit (j).
Table 2.
Means and 95% confidence intervals for the percent of papers audited and costs per paper.
Fig 3.
Mean false positive probabilities for 100 labs over time (research cycle) for the sensitivity analyses.
j = 75 cycle per audit (Auditing 1.35% of all papers) is the baseline scenario with: starting effort e0 = 75; increased effort after auditing el+ = en+ = 5; percentile cut-off for lab removal c = 0.67; minimum papers n = 50; review error = 0%. All other results change one of these six parameters. The baseline scenario is shown as a dashed line in all the panels bar the top-left panel where it is a solid line.
Table 3.
Percent of simulations that did not experience the competitive spiral.
Results from 16 scenarios. Results ordered by best (highest percent) to worst performing scenarios. See Table 1 for the key to the parameters.
Fig 4.
Means of effort, false positive and mistake probabilities for 100 labs over time (research cycle) with and without audits based on finding mistakes in published papers.
Table 4.
Means and 95% confidence intervals for the number of false positives and number of mistakes per 100 papers.
Results for two simulations, one without audits and one with audits with j = 200, el+ = 1 and en+ = 0.5.