Statistically Insignificant - Episode 8 (Podcast)
In episode 8 of the Statistically Insignificant podcast Jaryd and I are joined by my good friend Eric Mackay, a Senior Statistician at Cytel. Eric has an academic background in econometrics and statistics, as well as extensive professional expertise in applied biostatistical analysis. We cover a variety of topics including why Eric left the economics profession, the difficulty of translating statistical concepts into practical terms, and whether international travel is overrated. Note: this episode contains the use of technical jargon which some listeners may find boring.
- The likelihood of approval for oncology drugs is only 5%, the lowest success rate for all medical indications
- Vinay Prasad’s critique of surrogate endpoints and cancer drug efficacy
- Most cancer drug approvals are not based on overall survival, and the majority of associated clinical studies are at high risk of statistical bias
- Statistical paradises and paradoxes of big data: 400 truly random samples can be better than 2.3 million observations if the latter have a measurement bias
- Machine learning algorithms can erroneously conclude that asthma is a “protective” factor for pneumonia due to observational confounding
- For books introducing Bayesian statistics see: The Theory That Would Not Die and Bayesian Data Analysis