How likely is it that this pill or procedure will actually help me?
Healthy Skepticism is a short course for patients who want to make sense of medical information. It builds the intuitions behind the numbers you hear in the news and in the exam room, so you can have better, more confident conversations with your doctor. No prior statistics required, only curiosity and a willingness to ask "compared to what?"
Presented by Murray Cantor, PhD (mathematician, retired IBM Distinguished Engineer), assisted by Joel Keenan, MD (practicing physician).
One trial, two true descriptions: "risk fell by a third" and "2 people in 100 helped." (Illustrative numbers.)
Fall 2026
Healthy Skepticism OLLI at Yavapai College, a four-part series
Schedule
Thursdays, September 3 to 24, 2026
Weekly, 10:00 to 11:15 a.m.
Location
Yavapai College, Verde Valley Campus, Clarkdale, Arizona
Format
In person
Winter 2027
Healthy Skepticism: Making Sense of Medical Claims Community Library Sedona, a four-part series
Schedule
Thursdays, February 4 to 25, 2027
Weekly, 2:00 to 3:30 p.m. Arizona time
Location
Multi-Purpose Room, Community Library Sedona
Format
In person, adult audience
Four Sessions, Five Faults
Each session follows the same rhythm: short Khan Academy-style videos that start from a discrete case and generalize, questions posed before answers are given, worked examples, and a whitepaper that documents every source.
How Likely Is It That This Treatment Will Actually Help Me?
"Cuts heart attacks by a third" and "helps about 11 people in 1,000" can describe the same trial. Day one is Fault 1, relative reporting: the habit of asking what a headline percentage is a percentage of, and reading the absolute numbers underneath it.
The fault: relative reporting · Worked example: the Lipitor "36%" headline, shown as 1,000 people
What Does a Positive Test Really Tell You?
The probability of the evidence given the condition is not the probability of the condition given the evidence. Day two is Fault 2, probability reversal: how this quiet flip misreads screening results, and how to run the reasoning in the right direction.
The fault: probability reversal, not confusing P(A|B) with P(B|A) · Worked examples: a screening mammogram · a COVID rapid test · courtrooms and classrooms
Where "High" Begins, and Why the Advice Keeps Changing
Day three carries two faults. Fault 3, arbitrary categories: a cutoff turns a measurement into a label, and moving the line reclassifies millions of people whose numbers never changed. Fault 4, the Bernoulli fallacy: reading the chance of the data if there were no effect as the chance there is no effect, which is how single studies get overturned.
The faults: arbitrary categories · the Bernoulli fallacy · Worked examples: blood-pressure and BMI cutoffs · what a p-value does and does not say
When Does a Correlation Mean a Cause?
Two things move together in three ways without the first causing the second: coincidence, a shared cause, or the arrow running backward. Day four is Fault 5, the causal leap, and the whole day comes down to asking which diagram a claim assumes and what would rule the others out.
The fault: the causal leap · Worked examples: Nicolas Cage films & drownings · statins & car crashes · bird watching & dementia · and the one real cause, smoking & cancer
Whitepaper Library
You don't need any of these to get the full value of the course. The sessions are built to stand on their own. The whitepapers are here for anyone who finds the underlying analysis interesting, with the derivations worked out, and sources named.
Relative Risk, Absolute Risk, and the Arithmetic of Persuasion
Why one trial can be reported as a 95% cut or as a drop of less than a percentage point, worked through atorvastatin and the COVID-19 vaccines.
Sources: CTT Collaboration, ASCOT-LLA, CARDS, TNT, Olliaro et al. 2021
Absolute Mortality Risk as a Function of BMI
How the course's BMI risk curves are built, and where each assumption is most likely to mislead.
Sources: Global BMI Mortality Collaboration 2016, SSA period life table 2019, WHO TRS 894
Blood Pressure and Mortality
The dose-response between usual pressure and vascular death, and a Bayesian network running from age, sex, and both pressures to ten-year cardiovascular death.
Sources: Lewington et al. 2002, SPRINT 2015 and 2021, Franklin et al. 1999 and 2001, SHEP 1991
How Sure Is Sure Enough?
Why a finding that makes headlines in medicine would not count as a discovery in physics.
Sources: ATLAS and CMS 2012, Berger and Sellke 1987, Ioannidis 2005, ASA statement 2016
Student Materials
Students have access to the full slide decks, extended whitepapers, datasets, and session videos. The student page is password protected. The password is given out in class.