How to use this glossary
MiraProof articles often use terms borrowed from the research world — meta-analysis, RCT, confidence interval, and so on. You do not need to memorize them. When a word slows you down in an article, come here, read the everyday analogy, and go back. That is the fastest way to build a working vocabulary.
The terms are grouped into four buckets: study designs, study quality, measuring effects, and reading results. Read in order or skip around — whichever helps.
Study designs (how the research was done)
Meta-analysis
A statistical method that pools the results of many studies on the same question into a single combined estimate. Individually small trials can, together, reveal a stable overall pattern.
Everyday analogy: Asking ten doctors for a second opinion and then quantitatively combining their answers. Ten averaged opinions are usually less biased than any single one.
Caveat: If the source studies are low-quality, the pooled conclusion is also unreliable. Garbage in, garbage out.
Systematic review
A comprehensive overview that follows a documented protocol to find, screen, and summarize every study on a given question. Often paired with a meta-analysis, though not always with statistical pooling.
Everyday analogy: Sweeping through a library shelf by shelf, cataloguing every book on a topic and then producing a structured summary — nothing missed on purpose.
Randomized controlled trial (RCT)
A trial in which participants are assigned by lottery (randomly) to either receive an intervention or serve as a control, and outcomes are compared. RCTs are considered the "gold standard" for establishing cause and effect.
Everyday analogy: Splitting 100 cooking-class attendees into two groups by coin flip, having one group cook a new recipe and the other stick with the old one, then comparing taste scores. Because the starting groups were equivalent, any difference at the end is more plausibly the recipe's doing.
Why it matters: Randomization prevents the kind of bias where "healthier people happened to pick the new pill, so it only looked effective."
Double-blind
An RCT design in which neither the participants nor the researchers know who received the intervention and who received the placebo. Prevents expectations from swaying reporting or measurement.
Everyday analogy: A blindfolded judge tasting two unlabeled wines. Preconceptions get much harder to smuggle into the result.
Placebo
An inert pill or treatment designed to look identical to the real one. RCTs compare "real vs placebo" to subtract out the "belief effect" (placebo effect) from any measured benefit.
Everyday analogy: Two identical-looking mints in a bowl — one is a real mint, the other a flavorless sugar disc. The sugar disc is the placebo.
Cohort study
A study that follows a large group of people over years or decades, recording what health outcomes appear over time. Observational — no intervention.
Everyday analogy: Following 1,000 high-school classmates for 30 years and tracking how lifestyle differences relate to who develops which conditions. Great for spotting long-term patterns; weaker for proving cause.
Case-control study
A design that starts with people who already have the outcome (cases) and matches them with people who don't (controls), then looks backward for differences in exposure or behavior.
Everyday analogy: Comparing 100 people who had a traffic accident with 100 who didn't, asking each about sleep the week before, and looking for a pattern.
Cross-sectional study
A snapshot at a single point in time — measures health status and lifestyle in a population all at once. Good for "these patterns co-occur" but weak for "which caused which."
Everyday analogy: A single school health-check day that collects height and grades at the same time. You can see whether they correlate — you can't tell which is the cause.
Case report
A detailed write-up of a single (or small number of) patient(s). Useful for generating hypotheses; not enough to say a treatment works in general.
Everyday analogy: A doctor carefully documenting "my grandmother took X and got better." A starting point, not a conclusion.
In vitro / animal studies
Research done in petri dishes (cells only) or in animals such as mice. Important preliminary work, but effects seen in vitro or in animals do not always translate to humans.
Everyday analogy: Simulating a hurricane with a miniature model, then applying the result to a real city. Useful for direction; not a substitute for the real thing.
Study quality (how much to trust it)
Evidence hierarchy (evidence pyramid)
Study designs are ranked, roughly, by how directly they let you draw causal conclusions:
- Meta-analyses and systematic reviews
- Randomized controlled trials (RCTs)
- Cohort studies
- Case-control and cross-sectional studies
- Case reports and expert opinion
- In vitro / animal experiments
Higher on the ladder generally means stronger causal claims are possible. But "higher = always better" is a shortcut: a well-run cohort can be more trustworthy than a poorly run RCT.
Sample size
The number of participants in a study. Larger samples reduce the risk of being fooled by random noise.
Everyday analogy: Rolling a die six times and getting three sixes is not remarkable. Rolling it six thousand times and getting half sixes obviously is. Bigger samples let luck wash out.
Caveat: Bigger is not enough on its own — who is in the sample matters equally.
Population
The characteristics of the people in the study — age, sex, health status, country, lifestyle. An effect in "20 healthy young men" does not automatically apply to "women in their 50s."
Everyday analogy: A training regimen designed for professional athletes cannot be transplanted straight onto a sedentary grandparent.
Conflicts of interest (COI)
Financial or professional ties that could shape how results are interpreted. Examples: a drug company funding the trial for its own product, or an author advising a supplement brand.
How to read it: COI does not invalidate a study on its own. It's a cue to check whether the direction of the results conveniently favors the funder.
Reproducibility
Whether another team, running the same experiment under the same conditions, gets the same result. A single dramatic finding is weaker evidence than a modest finding replicated three times.
Everyday analogy: A recipe that works once in one kitchen might be luck. A recipe that reproduces across kitchens is a real recipe.
Measuring effects
Effect size
The magnitude of the observed effect, expressed numerically. Examples: "sleep duration increased by 15 minutes on average," "pain reduced by 30 %." A statistically real effect can still be too small to matter in daily life.
Everyday analogy: Shaving one second off a marathon is technically an improvement, but no one will notice in daily life. "Real" and "meaningful" are different questions.
Statistical significance (p-value)
The probability that a result this large (or larger) could have arisen by pure chance if there were no real effect. p < 0.05 (a 5 % chance) is the conventional cutoff for "unlikely to be pure noise."
Everyday analogy: If you roll a die and get twenty sixes in a row, "chance alone" starts to feel implausible. The p-value quantifies that "implausibility." Note: "significant" does not mean "large" — a tiny effect can be statistically significant in a huge study.
Caveat: Never read the p-value in isolation. Effect size and confidence interval belong in the same glance.
Confidence interval (CI)
A range of values within which the true effect is statistically likely to sit. Example: "weight loss averaged -2 kg (95 % CI: -3.5 to -0.5 kg)" means the true effect is most plausibly somewhere between -3.5 and -0.5 kg.
Everyday analogy: Similar to a weather forecast that says "tomorrow's high: 20 °C, plus or minus 3." The center point matters — but so does the width.
Reading tip: Narrower is more precise. When the interval crosses zero, the study cannot say with confidence that there is any effect at all.
Dose-response
The pattern where increasing the amount of exposure produces a proportionally larger effect. When present, this strengthens the case that the substance is really doing something.
Everyday analogy: One coffee versus three coffees — you feel more awake after three. If dose does not matter at all, the connection is probably weaker than it looks.
Reading MiraProof results (site-specific terms)
Confidence (weak / moderate / strong)
The three-bar ladder shown on every MiraProof article. It reflects study quality, quantity, and internal agreement combined. It answers "how much should I trust the current picture?" — not "does it work?"
Evidence direction
A summary of which way a body of research leans overall:
- Supportive: Studies consistently show benefit
- Mostly supportive: Mainly beneficial, with some exceptions
- Mixed: Positive and negative studies roughly cancel out
- Unsupportive: Studies consistently show no benefit
- Insufficient: Not enough good research yet to say either way
Verdict label
A one-word summary of evidence direction: "Effective," "Partial," "Limited," "Uncertain," "Not effective," and so on.
Caveat: The label is the doorway, not the answer. Before acting on it, check the article body for population, dose, effect size, and the confidence interval.
Where this fits
Nobody expects you to memorize these terms. Read your first evidence article, come back here for the two or three words that stopped you, and go on. Over a handful of articles, you'll pick up the vocabulary naturally.
To go one layer deeper, What counts as evidence? explains what "having evidence" really means, and How to judge study quality walks through four practical lenses for weighing a study.
This article is not medical advice. For personal health decisions, consult a physician or pharmacist.