The Independent Institute hosts their second livestream session of the weekend with S. Stanley Young, James E. Enstrom and moderator David Randall.
Well over 80% of claims made in observational studies fail to replicate. Some think meta-analysis studies using results from multiple observational studies are suitable for decision making. About 5,000 meta-analysis studies are published each year. I evaluate the reliability of meta-analyses in environmental epidemiology. I evaluate the base studies for p-hacking, trying this and that analysis to get a p-value less than 0.05. Plotting a p-value from each base study, ranked, against the integers, often shows a pattern consistent with p-hacking. I show that the base studies and the resulting meta-analysis are frequently unreliable, and have no sound statistical basis.
James E. Enstrom, Retired UCLA Research Professor and President, Scientific Integrity Institute:
“Reproducibility is Essential to Combating Environmental Lysenkoism”
My 2017 independent reanalysis of the ACS Cancer Prevention Study data found no robust relationship between fine particulate matter (PM2.5) and total mortality, contrary to the seminal 1995 Pope analysis that justified the 1997 EPA PM2.5 National Ambient Air Quality Standard. My reanalysis increases the evidence that inhaling about 100 micrograms of invisible fine particles per day does not cause premature deaths. In addition, it demonstrates the importance of access to underlying data in order to strengthen transparency and reproducibility in regulatory science. Finally, it increases efforts to combat the current form of Lysenkoism that has corrupted environmental epidemiology