Commenting on the study purporting to show falling rates of autism after the removal of thimerosal from infant vaccines, a reader who understands statistics far better than I do writes:
The statistical analysis in the thimerosal study you linked to isn’t even third rate. Interrupted time series analyses are notoriously unreliable in epidemiological research. But this article takes interrupted time series data and doesn’t even attempt a proper interrupted time series analysis.
The Danish study remains the most important and credible research on this subject.
Every child born in Denmark since 1968 has been given a national ID #, which the researchers could link with medical records. With a national health system, all vaccines were given through a single program, using a single vaccine manufacturer, so there was a relatively little variation in the pre/post thimerosal-removal date.
Moreover, the vaccine program had a standardized 3 dose sequence, an
additional consistency you wouldn’t find in the U.S. This enabled the
researchers to distinguish children who had received 0, 1, 2, or 3
thimerosal-containing doses, all within a narrow period around the
thimerosal-removal date. They found no dose-response relationship.
Others write in with more and more evidence that the Journal of American Physicians and Surgeons isn’t to be taken seriously. (See, for example, this editorial from Dr. (!) Grover Norquist.)
All of that said, I’d still like to see a serious analysis of the age-cohort data.