Sign up to save your podcasts
Or
Share GPH 42: Statistical Methods
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
GPH 42: Statistical Methods
In this episode, Medlock Holmes enters the engine room of epidemiology: statistical reasoning. Behind every risk ratio, confidence interval, and p-value lies a framework of probability, estimation, and uncertainty.
Holmes begins by revisiting foundational concepts - populations and samples, distributions, measures of central tendency, and variability. We explore how public health moves from descriptive statistics to inferential methods, allowing us to draw conclusions about populations from observed data.
The episode then examines key analytical tools used across epidemiology:
* Hypothesis testing and interpretation of p-values
* Confidence intervals and estimation precision
* Measures of association: risk ratios, odds ratios, rate ratios
* Regression models - linear, logistic, and Poisson
* Adjustment for confounding
* Interaction and effect modification
Holmes emphasises that statistics is not about mechanical computation, but disciplined interpretation. A statistically significant result is not necessarily clinically important. A non-significant result is not necessarily null.
We also explore model assumptions, overfitting, multicollinearity, and the importance of choosing the correct analytic framework for the research question.
Ultimately, statistical methods provide the grammar of public health evidence. Without statistical literacy, interpretation falters. With it, uncertainty becomes measurable and decisions become defensible.
Key Takeaways
* Statistics allows inference from samples to populations.
* Confidence intervals provide more insight than p-values alone.
* Measures of association must match study design.
* Regression modelling enables adjustment for confounding.
* Model assumptions must be checked to avoid misleading conclusions.
* Statistical significance and practical significance are distinct.
* Sound interpretation requires conceptual clarity, not just calculation.
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit drmanaankarray.substack.com/subscribe
View all episodes
By Med School Audio - Medical Knowledge Reimagined & Learning Made Memorable.In this episode, Medlock Holmes enters the engine room of epidemiology: statistical reasoning. Behind every risk ratio, confidence interval, and p-value lies a framework of probability, estimation, and uncertainty.
Holmes begins by revisiting foundational concepts - populations and samples, distributions, measures of central tendency, and variability. We explore how public health moves from descriptive statistics to inferential methods, allowing us to draw conclusions about populations from observed data.
The episode then examines key analytical tools used across epidemiology:
* Hypothesis testing and interpretation of p-values
* Confidence intervals and estimation precision
* Measures of association: risk ratios, odds ratios, rate ratios
* Regression models - linear, logistic, and Poisson
* Adjustment for confounding
* Interaction and effect modification
Holmes emphasises that statistics is not about mechanical computation, but disciplined interpretation. A statistically significant result is not necessarily clinically important. A non-significant result is not necessarily null.
We also explore model assumptions, overfitting, multicollinearity, and the importance of choosing the correct analytic framework for the research question.
Ultimately, statistical methods provide the grammar of public health evidence. Without statistical literacy, interpretation falters. With it, uncertainty becomes measurable and decisions become defensible.
Key Takeaways
* Statistics allows inference from samples to populations.
* Confidence intervals provide more insight than p-values alone.
* Measures of association must match study design.
* Regression modelling enables adjustment for confounding.
* Model assumptions must be checked to avoid misleading conclusions.
* Statistical significance and practical significance are distinct.
* Sound interpretation requires conceptual clarity, not just calculation.
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit drmanaankarray.substack.com/subscribe