Science1 distinct publisher3 min readUpdated
A PLOS One survey of 1,000 adults puts regular statistics users at 11%. If you publish risk numbers, forecasts or dashboards, that is your readership.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
The published shares carry more information than the headline. A quarter of the 1,000 respondents said they understood nothing about statistics [4][5]; 37% said their familiarity was limited [6]; a little over a quarter said they had learned some at school [7]; and the rest reported regular use [3]. Those four shares add to roughly 99% [2], which is what you get from the rungs of a single self-report ladder rather than four independent measurements. Read that way, the much-quoted 62% is the bottom two rungs added together [1], and the regular-user figure is the top rung of the same item, not a separate count of people who actually work with data [2].
What was asked matters more than how it was totalled. The item paired two different things in one sentence: "How much do you understand about statistics and p-values?" [8]. Someone who follows a weather forecast or a batting average without difficulty, but has never had cause to interpret a p-value, has an honest reason to answer "limited". The 62% is the share who decline to vouch for themselves on the harder half of a compound question.
The paper says as much in its title, which is about self-reported perception [15]. Two self-assessment questions were analysed and no ability test was administered [8], so the study cannot say how wrong people are about themselves. For anyone shipping a risk estimate, perception is the operative variable anyway. A reader who believes he cannot evaluate an interval will read it as "they are not sure" and move on.
The scale is worth stating plainly. For every adult who reports regular use there are about 5.6 who report no or limited knowledge [3], so the regular user is roughly one reader in nine [4]. Sampling noise does not rescue this: at 1,000 respondents, a proportion near half carries a 95% interval of about three points either way [5]. The ordering is stable even if the exact percentages are not.
The appetite finding invites a matching discount. Agreeing that you would lean on statistics more if you understood them better costs the respondent nothing, and Ramos, who led the study, notes that free material already exists, including Penn State's STAT 200 course pages, posted for people who are not enrolled [14]. Supply is not the binding constraint. A two-question instrument also says nothing about which presentations land, so this survey identifies an audience rather than a remedy.
"While statistics are not hard to understand, they are even easier to misunderstand," Ramos said [11]. His other line is the one to keep near a dashboard: statistics does not deliver certainty, it works with the uncertainty already present [13]. Readers who report they cannot follow the notation will not learn that from the notation, and it is the notation that most public evidence still travels in [12].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
A study published in PLOS One and led by Mark Ramos, assistant research professor of health policy and administration at Penn State, analysed nationally representative survey data and found 62% of respondents reported no or limited statistical knowledge.
The survey covered 1,000 people.
One-quarter of those surveyed reported that they had no understanding of statistics.
37% reported limited familiarity with statistics.
Slightly more than one-quarter said they had learned some statistics in school.
Ramos and Anyaso-Samuel submitted two questions that were used in the survey, "How much do you understand about statistics and p-values?" and "How often would you base decisions on reported statistics if you understood it better?", and analysed the responses to those two questions for the study.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Peer-reviewed but single self-report instrument
The finding rests on a published PLOS One paper with a DOI, a nationally representative sample of 1,000 US adults, and internally consistent reported shares (the four categories total about 99%). Against that: the measure is two piggyback questions on a commercial omnibus survey, all outcomes are self-assessed perception rather than tested knowledge, the survey vendor is unnamed, and no weighting, response rate or margin of error is disclosed — precision has to be inferred from n alone.
Low measured everyday use, high stated appetite
Adoption here is the real-world use of statistics by the audience being measured, and the survey discloses it directly: 11% report regular use in daily life, roughly 89% do not, and 62% self-place in the lowest two knowledge bands. The offsetting signal is stated intent — nine in ten say they would use statistics more with better understanding — but intent is not usage, so measured adoption stays low.
Perception reported as capability
The framing runs slightly ahead of the instrument. The headline says adults 'lack basic statistical understanding' and the cluster titles it a '62% problem', but the underlying paper is scoped to 'self-reported perception of statistical literacy' and the 62% is an arithmetic sum of two self-assessment bands with no objective test, no disclosed weighting and no margin of error. The direction of the finding is plausible and the study is peer reviewed, so this is overstatement of certainty rather than fabrication; the coverage also avoids policy or product claims it cannot support.
University communications channel with a course to promote
The only source is a research-institution write-up republished by an aggregator: the lead author is a Penn State research professor, the framing is favourable to the field of statistics, and the article closes by promoting Penn State's own STAT 200 course materials as the remedy to the problem it has just described. That is a normal academic-publicity incentive rather than a commercial one, and the presence of a DOI-identified peer-reviewed paper constrains it, but with no independent outlet in the cluster nothing checks the framing.
Verifiable paper, unverified methodology, one publisher
Confidence is moderate. The underlying study is identifiable and peer reviewed, the internal arithmetic checks out, and the numbers are quoted consistently — but the cluster has exactly one publisher and one article, the survey vendor and weighting are undisclosed, and every figure is self-reported. Enough to cite the direction and rough magnitude; not enough to treat the specific shares as precise population estimates.
science
A vaccine communicator's testable claim: hesitancy answers to burden, not to better facts1 distinct publisher
product
How a short GNET essay on AI slop became a settled fact about children's brains1 distinct publisher
science
Transcription caught mid-act in fly embryos, and it does not match the test tube1 distinct publisher
science
Mitchell's complaint: we grade models with human exams when the control group is animals and infants1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
phys.org
1 article · August 21, 2026