Once again, just go back to the old school ways. These new ways are not robust at all. You need to regularly perform independent replications (no, it shouldn't

Once again, just go back to the old school ways. These new ways are not robust at all. You need to regularly perform independent replications (no, it shouldn't be done in parallel and published in the same paper). This will filter fraud, honest errors, and non-stationary effects. It will also encourage publication of usable methods sections, and much more.