BioSuite AI PortfolioEvidence & DiagnosticsStatEvidence
StatEvidence: Understand the statistics behind a result. Explore statistical methods, assumptions and interpretation. Interactive demo.
StatEvidenceExplore statistical methods, assumptions and interpretation.Understand the statistics behind a result. · Interactive demo
PREM KUMARSCIENCE · LEADERSHIP · AI
← All appsStatEvidence DEMO VERSION

BIOSTATISTICS · UNDERSTAND THE UNCERTAINTY

Beyond the average.
See the evidence.

Explore measurements, compare two independent groups and see how study assumptions change the sample size.

Educational tool by Prem Kumar Perumal. Use fictional or de-identified numeric measurements only. Data stays in your browser. This prototype is not validated for clinical or regulatory decisions.

01 / MEASUREMENTS TO EVIDENCE

Two groups. One useful question.

Are the group means different, and how uncertain is the estimated difference?

Comma, space or new-line separated. Up to 5,000 values per group; same units in both groups. Missing values and text are rejected.

Inspect every measurement.

Group measurements and mean confidence intervals

Each dot is a measurement; overlapping dots can hide duplicates. Gold diamonds show the means; horizontal lines show 95% confidence intervals for the means, not ranges of individual measurements. All accepted values are plotted.

Full descriptive statistics

02 / PLAN BEFORE COLLECTING

How many observations?

Explore an approximate sample size for detecting a specified mean difference with two independent groups.

Two-sided α = 0.05. Equal allocation and common SD. SD and difference must use the same units.

A planning estimate, not an exact t-test power calculation. Small samples, unequal variances, paired observations, clustering and repeated measures need a design-specific method. Attrition inflation assumes the remaining observations still satisfy the design assumptions.

03 / TRANSPARENT METHODS

The method matters.

Descriptive statistics use the sample SD (n−1 denominator). Mean intervals use Student’s t with n−1 degrees of freedom. Comparison uses a two-sided Welch t-test of a zero mean difference, with Welch–Satterthwaite degrees of freedom and a 95% interval for A−B. Probabilities are calculated using the t distribution.

Observations must be independent, within and between groups. For small samples the groups should be approximately normal; inspect outliers and the distribution. This app does not test normality, handle paired data, adjust for confounding or correct for multiple tests. A small p-value does not measure effect size, establish causation or make a difference clinically important. A large p-value does not establish equivalence.

A 95% confidence procedure covers the fixed population mean or difference in about 95% of repeated samples under its assumptions. If both groups have zero variance, the test and difference interval are withheld; collapsed descriptive intervals are flagged.

Sample-size formula & sources

Analyzable n per group ≈ 2 × (z₀.₉₇₅ + zpower)² × SD² / difference², rounded up with a minimum of 2. Recruitment n = ceiling(analyzable n / (1 − attrition)). This is the normal approximation for two means.