Sample Size of One
Asystole

Welcome to journal club.
MC Anecdote versus Doctor Data.
Keep it clean. Cite your sources.

My mom says I’m the greatest.

Conflict of interest.

You crowned yourself the greatest from a poll inside your phone;
Your sample was your fan club, and you filled it out alone.
I set my alpha early, before the numbers hit;
You picked yours after peeking. What a tailored little fit.

Under the null, with model assumptions in view,
Results this extreme or more — how often come through?
That probability’s p. You flipped the question to
“What’s the chance the null is true?” That’s a different thing to do.

Your p got small and you started acting different:
First time something that small got called significant.
Below the preset alpha gets that label on the page;
A tiny effect can earn it. Check the size before the praise.

Big p? You yell, “No difference!” Slow that victory speech;
Low power can leave real effects beyond your study’s reach.
Twenty tests, one lucky hit — you put it on display;
I show the whole family. You hide nineteen away.

T-test for means, chi-square for counts;
Check your assumptions before your announcements.
Show the effect and uncertainty too —
Your ego’s got a sample size of you.

You said, “I’m mean.” Congratulations on the pun.
A t-test checks mean differences. Let’s actually run one.
Separate groups? Independent observations on the sheet;
Welch handles unequal variances. Now stay on beat.

Same people, pre and post? Keep every person paired;
Test the mean of their changes. That’s the gap compared.
Check the shape and outliers before you make the call;
You skimmed one abstract, now you’re teaching in the hall.

Chi-square takes categories, counts in every square;
Under independence, what’s expected there?
Observed minus expected, square it, then divide
By expected; sum the cells. That’s the score supplied.

Tiny expected counts in a two-by-two display?
Fisher’s exact can have the final say.
I check assumptions while you’re flexing in the booth;
Your spreadsheet wore a lab coat. You mistook it for the truth.

T-test for means, chi-square for counts;
Check your assumptions before your announcements.
Show the effect and uncertainty too —
Your ego’s got a sample size of you.

Incidence: new cases, with the period defined;
Incidence rate puts person-time at risk beneath the line.
Point prevalence: existing cases over population;
A snapshot, at one time. That’s the whole calculation.

Risk ratio: risk exposed over risk without;
Odds ratio: odds compared. Mind what leaves your mouth.
Odds: chance it happens over chance it doesn’t — check;
Odds and risk can differ. Keep your ratios correct.

Disease risk drops from two percent to one in the same span;
“Fifty percent reduction!” You’re your own publicity man.
One percentage point absolute. A hundred treated here
Prevent one case on average. See the baseline reappear.

Coffee tracks with cancer? Check who smokes in either crew;
A confounder can distort the link you’re pointing to.
Association needs more work before causation takes the crown;
Your only control group is the friends you boss around.

Type one: true null, wrongly rejected.
Type two: false null, left undetected.
Power is one minus beta, for the effect in view;
A bigger sample won’t erase the bias built into you.

Your confidence is huge; your interval is wide.
I show uncertainty. You keep yours classified.
You said your work was peer-reviewed. I checked the situation:
Your roommate saying “looks good” isn’t external validation.

T-test for means, chi-square for counts;
Check your assumptions before your announcements.
Show the effect and uncertainty too —
Your ego’s got a sample size of you.

All right. Rematch?

Sure. Preregister.

Ooooooh!

Then replicate.