Mathematics graduate working on statistical inference in epidemic models β specifically on how much of a published conclusion survives the choices its author never wrote down.
Independent researcher, Lahore, Pakistan. Currently applying for funded Master's study in applied / computational mathematics.
dengue-multiverse β Does the evidence for climate-driven dengue transmission survive the analyst?
221 dengue outbreaks across 33 countries, each fitted under all 144 combinations of six routine analysis choices β 33,152 model fits. Roughly three-quarters of the variation in whether climate forcing is endorsed lies between analyses of the same outbreak, not between outbreaks. Most of that is an outbreak Γγ°γ€ analysis interaction that no standardised convention can remove.
The paper does not stop at the diagnosis: combining eight analyses by Rubin's rules turns a 95% confidence interval that covers 71% of the time into one that covers 97%, for a few minutes of computation.
Python Β· SciPy Β· compartmental models Β· maximum likelihood Β· multiverse analysis
I keep a dated analysis log in every project recording the formulations that failed, the hypotheses that were tested and rejected, and the bugs β including the ones that produced plausible wrong numbers rather than crashing. A result is easier to trust when the route to it is visible, including the wrong turns.
Where a paper quotes a number, a test checks that number against the pipeline that produced it. If the manuscript drifts, the test suite fails.
Python NumPy SciPy pandas Matplotlib LaTeX MATLAB SQL pytest
Background: BS Mathematics (University of the Punjab) Β· 3.5 years teaching mathematics at college level Β· data analyst
Interests: mathematical epidemiology Β· numerical methods for differential equations Β· model selection and identifiability Β· reproducible research
Corrections and objections to anything here are welcome β open an issue.