Can someone who takes the plain to explain this kind of things " For the uninitiated, there are major risks associated with reliance on large, raw (uncorrected) data sets for retrospective (backwards in time) data sets.  The key technical term here is "confounding variables", but data entry errors (such as multiple entries for the same diagnostic event) or process changes can also introduce huge sources of bias into large data sets like this.  With raw data, it is most useful to consider any data plotting to be sort of a first draft, useful for identifying potential trends or topics that deserve more detailed analysis.  But sometimes, when the observed effect size in the raw data is very large or potentially important, alarm bells start ringing even before full analysis is completed.  And that seems to be the case with these data."
be just a conspiracy theorist?
If not, we'd better be concern with his analysis of some raw data dumped by some US army whistleblowers:
https://rwmalonemd.substack.com/p/regarding-the-defense-medical-epidemiological