Ananya Iyer

Explain a Statistical Result in Plain English, Honestly

Translate a study result into plain English: every statistic defined, relative risk turned into absolute numbers, and common misreadings corrected.

At a glance

Best for: Writers, managers, students and health or wellbeing teams passing on a study finding to people who won't read the paper.
Tested on: Claude · Opus 5.5
You fill in: audienceresultdesign
You get: Takeaway In this study, people who meditated with an app for 10 minutes a day were somewhat less likely to report poor sleep six months later, but… (full result below)

Prompt

Explain this statistical result to {{audience}} in plain English, without overstating it.

The result, exactly as reported: {{result}}
Study design (as much as I know): {{design}}

Structure:

1. A one-sentence takeaway a non-specialist could repeat accurately.

1. What each number means: define every statistic in the result (p-value, confidence interval, risk ratio, effect size, etc.) in one plain sentence each, using this study's numbers, not a textbook example.

1. Absolute vs relative: if the result is relative (a % change or a ratio), convert it to an absolute difference using the baseline, or tell me what baseline I'd need to find.

1. What it does NOT mean: the 3 most likely misreadings, each corrected.

1. How much weight to put on it: given the design, is this correlation or causation, and what would a stronger study look like?

Rules: no jargon without an immediate definition. Don't say "statistically significant" without saying what it does and doesn't imply. If the reported numbers look internally inconsistent, point it out.

Fill in

{{audience}}: Who the explanation is fore.g. readers of a company wellbeing newsletter
{{result}}: The result, copied exactly, including all numberse.g. 23% lower risk of poor sleep (RR 0.77, 95% CI 0.61-0.97, p = 0.03)
{{design}}: What you know about how the study was donee.g. Observational cohort, 2,400 UK adults, self-reported app use and sleep

Example input

audience: readers of a company wellbeing newsletter
result: Adults who used a meditation app for at least 10 minutes a day had a 23% lower risk of reporting poor sleep after six months (risk ratio 0.77, 95% CI 0.61 to 0.97, p = 0.03). Among non-users, 31% reported poor sleep.
design: Observational cohort of 2,400 UK adults. App use was self-reported at the start; sleep quality was self-reported by survey after 6 months. No random assignment.

1.3k0

Generated Outputs (1)

3 hours ago

Claude

Opus 5.5

3 hours ago

Generated Output

1. Takeaway
In this study, people who meditated with an app for 10 minutes a day were somewhat less likely to report poor sleep six months later, but the study can't show that the app caused it.

2. What the numbers mean

• Risk ratio 0.77: app users reported poor sleep 0.77 times as often as non-users. That's where "23% lower" comes from (1 - 0.77).

• 95% confidence interval, 0.61 to 0.97: the range of values reasonably consistent with the data. The true effect could be anywhere from a large reduction (39%) to a tiny one (3%).

• p = 0.03: if meditation apps made no difference at all, a gap this big or bigger would turn up by chance about 3% of the time. It's "statistically significant", which means chance alone is an unlikely explanation. It doesn't mean the effect is large, or that the app caused it.

3. Absolute numbers
Among non-users, 31% reported poor sleep. For app users, that's about 24% (31% x 0.77). So the absolute difference is about 7 percentage points: roughly 7 fewer people in every 100 reporting poor sleep. Put another way, about 14 people would need to meditate daily for one fewer person to report poor sleep, assuming the effect were causal.

4. What it doesn't mean

1. "Meditation cuts insomnia by 23%." No. The outcome is self-reported poor sleep, not diagnosed insomnia, and the result shows an association, not a cause.

1. "23% of people will sleep better." No. It's a relative figure; the absolute difference is about 7 in 100.

1. "It works for everyone." No. It's an average across a group, and says nothing about any one person.

5. How much weight to give it
Modest. It's observational, so people who meditate daily may also exercise more, drink less or have calmer lives, and any of those could explain better sleep. Both app use and sleep were self-reported, too. A stronger study would randomly assign people to use the app or not, and measure sleep more objectively.

Consistency check: the numbers fit together. A risk ratio of 0.77 sits in the middle of the interval on the log scale, and that interval matches p of about 0.03.

Why it works

"Using this study's numbers, not a textbook example" makes the definitions useful rather than generic. The absolute vs relative step is the single most important correction in health reporting: "23% lower risk" sounds large until it becomes "about 7 in 100 people". Asking for three misreadings anticipates how the result will be misquoted. The internal consistency check is an easy way to catch typos in reported statistics. Linking the weight of evidence to study design turns "significant" into an honest judgement.

When not to use it

It can only explain what you give it; it can't assess a study's quality without its methods, so read the paper or a trusted summary for anything you'll base decisions on. Not for medical advice to individuals. Complex models (multilevel, Bayesian, survival curves) need a statistician's reading, not a plain-English gloss.

Comments (0)

No comments yet. Be the first to share your thoughts!