Multilevel Power & MDES Calculator

1 Study design

Randomization structure

Choose the level where treatment is randomized; choosing the wrong structure can substantially misstate precision.

MDES from sample Sample from target

2 Error rates & allocation

Alpha

False-positive rate; setting it too high makes the study look more precise than the intended test.

Power

Chance of detecting the target effect; overstating it understates the sample required.

Test

Use one-tailed only when effects in the opposite direction would not count as evidence.

Proportion treated

Share assigned to treatment; imbalance away from .50 reduces precision.

3 Sample structure

4 Variance & covariates

Variance shares

Covariate adjustment

Do not guess ICCs. Source empirical values from published ICC compendia or closely matched prior data for the outcome, grade, and population; optimistic ICC or R² assumptions understate MDES.

5 Expected attrition

Cluster/site attrition

Share of assignment units expected to be lost; setting it too low overstates the analyzable sample.

Individual attrition

Share of people with missing outcomes; setting it too low overstates within-cluster precision.

Planning assumptions

Balanced cluster sizes and treatment allocation.
Continuous, approximately normally distributed outcome.
Two-arm comparison with correctly specified variance components.
Equal ICCs and residual variances across treatment arms.

Planning result

Two-level cluster random assignment

Minimum detectable effect size

—SD units

Calculated with finite-sample t critical values.

Planned sample

—

Enter assumptions to calculate.

t multiplier—

Analysis units—

Degrees of freedom

—

Finite-sample reference

Effective clusters / sites

—

After cluster attrition

Effective individuals

—

After all attrition

Design effect

—

Variance vs. equal-sized SRS

Sensitivity to number of clusters

All other assumptions held fixed

Current design

Sensitivity to cluster-level R²

All other assumptions held fixed

Current assumption

Formula used

Proposal-ready methods paragraph

Review the assumptions and edit wording to match your analysis plan.

Runs entirely in your browser. Nothing you type is sent anywhere or stored. All free tools →