Dunnett Test Calculator

Enter Study Data

Use raw observations or summary statistics. Leave unused treatment rows blank.

Responsive calculator grid: 3 columns large, 2 medium, 1 small

Input Mode

Raw observations

Summary statistics

Familywise Alpha

Alternative

Displayed Decimals

Simulation Draws

Random Seed

Control Group Name

Raw Observation Entry

Control Values

Separate numbers with commas, spaces, or line breaks.

Treatment 1 Name Treatment 1 Values

Treatment 2 Name Treatment 2 Values

Treatment 3 Name Treatment 3 Values

Treatment 4 Name Treatment 4 Values

Treatment 5 Name Treatment 5 Values

Treatment 6 Name Treatment 6 Values

Summary Statistic Entry

Control Summary

n Mean Standard Deviation

Treatment 1 Name n Mean Standard Deviation

Treatment 2 Name n Mean Standard Deviation

Treatment 3 Name n Mean Standard Deviation

Treatment 4 Name n Mean Standard Deviation

Treatment 5 Name n Mean Standard Deviation

Treatment 6 Name n Mean Standard Deviation

Formula Used

1) Pooled error variance

MSE = Σ[(n_i - 1)s_i²] / (N - g)

2) Dunnett comparison versus control

t_i = (x̄_i - x̄_c) / √[MSE(1/n_i + 1/n_c)]

3) Correlation between treatment contrasts

ρ_ij = (1 / n_c) / √[(1/n_i + 1/n_c)(1/n_j + 1/n_c)]

4) Familywise critical value

The calculator estimates the Dunnett critical limit by Monte Carlo simulation from the multivariate t distribution defined by the shared control and pooled error degrees of freedom.

How to Use This Calculator

Choose raw observations if you have individual values, or summary statistics if you only know n, mean, and standard deviation.
Enter the control group first. Then fill one or more treatment groups.
Set the familywise alpha level, alternative direction, simulation count, and displayed decimals.
Click Calculate Dunnett Test. Results appear above the form below the header.
Review ANOVA first, then inspect Dunnett adjusted p values, simultaneous confidence intervals, and final decisions.
Use the CSV and PDF buttons to export the descriptive tables, ANOVA summary, and Dunnett comparison results.

Example Data Table

Group	Observations	n	Mean	SD
Control	18, 20, 21, 19, 22	5	20.0	1.5811
Low Dose	24, 23, 22, 25, 24	5	23.6	1.1402
Medium Dose	20, 21, 19, 22, 20	5	20.4	1.1402
High Dose	27, 26, 25, 28, 27	5	26.6	1.1402

These values are prefilled in the default form so you can test the calculator immediately.

FAQs

1) What does the Dunnett test do?

It compares several treatment means against one control mean while controlling the familywise error rate. It is useful after one-way ANOVA when your interest is focused on control-versus-treatment contrasts.

2) When should I use this calculator?

Use it when you have one control group and multiple treatment groups under equal-variance assumptions. It is ideal for experiments, quality studies, and intervention testing where every comparison targets the same control.

3) Can I use raw data and summary data?

Yes. Raw mode accepts individual observations. Summary mode accepts sample size, mean, and standard deviation for each group. Both routes produce the same structure of pooled ANOVA and Dunnett comparisons.

4) Why is a pooled variance used?

Classical Dunnett testing assumes a common within-group variance. The pooled mean square error combines all group variances into one estimate, improving stability when equal-variance assumptions are reasonable.

5) What do the adjusted p values mean?

They are control-versus-treatment probabilities corrected for testing several treatment contrasts at once. A small adjusted p value indicates evidence against the null comparison after familywise error protection.

6) Why are simulation settings included?

The calculator estimates Dunnett critical values from the multivariate t distribution using Monte Carlo simulation. More draws usually improve stability, while the seed helps reproduce the same results later.

7) What assumptions should I check?

Check independence, roughly normal residuals, and similar variances across groups. Serious outliers or strong variance differences can distort pooled-error methods and should be reviewed before interpretation.

8) What does the simultaneous confidence interval show?

It gives a familywise-protected range for each treatment minus control mean difference. If a two-sided interval excludes zero, that comparison is significant at the chosen familywise alpha level.