Data Science x Public Health

Model averaging is often presented as a more careful and uncertainty-aware alternative to choosing one model specification. It is supposed to reduce overconfidence and make analysis more robust. But what if all the models being averaged share the same blind spots from the start? In this episode, we break down why model averaging often overpromises, how shared structural weaknesses survive the averaging process, and why uncertainty cannot be handled simply by blending similar models.&nbs...

Resource allocation models are supposed to help public health systems distribute scarce resources more intelligently. They promise better targeting, more efficient deployment, and stronger impact under constraint. But what if the model is optimizing inside a system whose deepest constraints should never have been treated as fixed? In this episode, we break down why resource allocation models often fail in practice, how optimization can normalize structural scarcity, and why better public he...

Competing risks methods are often presented as a more realistic way to analyze time-to-event data in epidemiology and public health. They promise to handle situations where other events prevent the outcome of interest from ever occurring. But what if the method becomes more sophisticated while the interpretation becomes less clear? In this episode, we break down why competing risks analyses are often overtrusted, how the choice of estimand quietly changes what the result means, and why ...

Adjustment for baseline differences is one of the most common moves in health research and biostatistics. It is often treated as proof that two groups have been made more comparable and that bias has been reduced. But what if that adjustment is creating more confidence than the data actually deserve? In this episode, we break down why baseline adjustment often fails, how observed balance can hide deeper structural non-comparability, and why adjusting for differences is not the same as s...

Public health scorecards are supposed to improve accountability and make system performance easier to track. They promise clarity, targets, and faster decision-making. But what if the scorecard starts changing behavior in the wrong direction? In this episode, we break down why public health scorecards often fail, how metrics become incentives, and why better-looking numbers can still hide weaker real-world performance. 👉 Enjoyed the episode? Follow the show to get new episodes automatically...

Secondary attack rates are often used to estimate how infection spreads among close contacts. They seem to provide a focused measure of transmission in households, schools, workplaces, and other settings. But what if the number is being shaped just as much by contact tracing and testing rules as by the pathogen itself? In this episode, we break down why secondary attack rates often mislead, how inconsistent contact definitions distort interpretation, and why outbreak metrics cannot be s...

Statistical power is one of the most familiar concepts in biostatistics and research design. It is supposed to help determine whether a study can detect a meaningful effect. But what if power is being used the wrong way after the study is already finished? In this episode, we break down what statistical power actually means, why post hoc power language often misleads, and why large or “well-powered” studies are not automatically strong evidence. 👉 Enjoyed the episode? Follow the show to...

Health equity dashboards are supposed to make disparities visible and drive better public health decisions. They promise transparency, accountability, and measurable progress. But what if the dashboard is making inequity easier to display without making it easier to solve? In this episode, we break down why health equity dashboards often fail, how they can turn structural inequality into performance theater, and why visibility is not the same as meaningful institutional change. 👉 Enjoyed t...

Statistical significance is one of the most familiar ideas in research. It is often treated as the dividing line between real evidence and random noise. But what if that binary framing is doing more harm than good? In this episode, we break down why statistical significance often misleads, how threshold thinking distorts interpretation, and why effect size, uncertainty, and design quality matter far more than a simple significant versus non-significant label. 👉 Enjoyed the episode? Follow the...

Predictive models are widely used to identify high-risk patients and populations. They promise earlier intervention, better resource allocation, and improved outcomes. But what if prediction alone is not enough to actually change what happens next? In this episode, we break down the critical difference between prediction and causation—and why models that perform well statistically can still fail when used in real-world decision-making. You will learn why predicting risk is not the same as k...

Case fatality rate is one of the most commonly cited numbers during outbreaks and health emergencies. It seems to offer a direct answer to a simple question: how deadly is this disease? But what if the rate is being shaped less by biology and more by who gets detected as a case? In this episode, we break down why case fatality rates often fail when detection is unequal, how testing and surveillance distort the denominator, and why context matters when interpreting outbreak severity. 👉 E...

Standard errors are one of the most overlooked pieces of statistical output. They sit underneath confidence intervals, p-values, and claims about precision in almost every study. But what if those standard errors are wrong from the start? In this episode, we break down what standard errors actually represent, why they often fail when real-world data violate model assumptions, and how this creates false confidence in research findings. 👉 Enjoyed the episode? Follow the show to get new ep...

Benchmark accuracy is one of the most trusted signals in machine learning. It tells you which model performs best—and it often drives decisions about what gets deployed. But what if that number is giving you a false sense of confidence? In this episode, we break down why models that perform well on benchmarks often fail in real-world settings. You will learn how dataset assumptions, evaluation metrics, and deployment conditions create a gap between leaderboard success and practical reli...