Things that apparently cause cancer
A replication study applies a Harvard nuclear plant cancer methodology to arbitrary landmarks like Costco, revealing that the flawed proximity-score model absurdly proves everything causes cancer.
read it The piece provides a sharp methodological teardown of flawed ecological epidemiology through an entertaining and revealing replication experiment.
summary
Researchers evaluated a recent Harvard methodology linking nuclear power plant proximity to increased cancer mortality by applying its exact spatial model to unrelated landmarks like Costco and sports stadiums. The statistical model generated massive false positives, attributing millions of cancer deaths to retail stores and universities with zero radioactive emissions. This experiment demonstrates that distance-weighted ecological regressions can output positive results for any coordinate cluster, making them useless for determining genuine environmental health risks. The critique underscores the necessity of establishing concrete exposure mechanisms and proper controls before publishing alarming public health claims.
key points
- The proximity-score model attributed over 2.2 million cancer deaths to Costco and 1.59 million to private universities using the exact logic applied to nuclear plants.
- Professional sports venues without emissions also yielded high statistical associations with cancer mortality under the model.
- The underlying methodology relies on broad ecological correlations over hundreds of kilometers without establishing biological exposure pathways or doses.
who should read it
Data scientists, researchers, and technical readers interested in the pitfalls of observational epidemiology.
discussion positive
consensus
Commenters thoroughly enjoy the brutal methodological takedown of sensationalist public health papers, though several technical readers note that the critics had to guess at the original code through trial and error.
pushback
A minority argue that ecological studies only aim to highlight population-level associations rather than direct causation, suggesting the critique oversimplifies the original research scope.
As someone with experience in this field (and plenty of skepticism of geographic association studies), I find the article lacking. Is there a technical write-up?
It is quite odd that the authors say that they don’t know if the original method can yield negative results. Some of the results are expressed as relative risk, so there’s a numerator and denominator. That should be enough to begin to work it out, since all possible exclusive numerators sum to the denominator. It feels like if you don’t understand whether some areas will be lower than average, then perhaps you don’t understand the …
It challenges the replication methodology, pointing out that reverse-engineering through trial and error might lead to an overfit model that does not fairly represent the original paper.
Generated by an LLM from the article and thread. Check the sources before relying on it.