Quick Answer: Dunnett’s Test Result

The Dunnett’s Test result shows that all three comparison groups are significantly higher than the control group. The control group is studytime group 1, with a mean G3 of about 10.84. Group 2 has a mean of about 12.09, group 3 has a mean of about 13.23, and group 4 has a mean of about 13.06.

The adjusted p-value chart shows that every treatment-versus-control comparison is below α = .05. The adjusted p-values are approximately 2.794e-05 for group 2 vs control, 2.422e-09 for group 3 vs control, and 0.0003436 for group 4 vs control. The confidence-interval chart also supports the same conclusion because the intervals for all treatment-control mean differences remain above zero.

What Is Dunnett’s Test?

Dunnett’s Test is a multiple-comparison procedure used after ANOVA when the research design has one control group and several treatment groups. It adjusts for multiple testing while keeping the analysis focused on the control comparison question.

This makes Dunnett’s Test different from all-pairs post hoc methods. A method such as Tukey, Bonferroni or Duncan may compare every possible pair, but Dunnett’s Test compares only each treatment group against the selected control. This is useful when the research question is not “which groups differ from each other?” but “which treatment groups differ from the control?”

In this example, studytime group 1 is treated as the control group. Groups 2, 3 and 4 are compared with group 1. The result shows that all three groups are significantly higher than the control group in mean G3 final grade.

This guide connects naturally with One Way ANOVA, ANOVA in SPSS, ANOVA in R, ANOVA in Python, ANOVA Assumptions, P Value, Confidence Interval, Effect Size, Eta Squared and Omega Squared.

Dunnett’s Test Formula

Dunnett’s Test begins with the difference between each treatment mean and the selected control mean:

The standard error for each treatment-control difference is commonly based on the ANOVA mean square error and the sample sizes:

The test statistic is:

Dunnett’s method then adjusts the decision for the fact that several treatment groups are being compared with the same control group. Software usually reports adjusted p-values and adjusted confidence intervals. A comparison is significant when the adjusted p-value is below the selected alpha level, usually .05, or when the adjusted confidence interval does not include zero.

Key Terms in the Formula

Dunnett’s Test Hypotheses

Dunnett’s Test uses one comparison for each treatment group against the control. In this example, group 1 is the control, so the hypotheses compare groups 2, 3 and 4 against group 1.

Dataset and Variables Used

The worked example uses student performance data. The dependent variable is G3 final grade. The grouping factor is studytime, which has four groups. For Dunnett’s Test, studytime group 1 is selected as the control group.

Group Mean Pattern

The mean pattern explains the Dunnett result. Every comparison group is above the control mean. Group 3 has the largest observed difference from control, while group 4 is also far above control but has the smallest sample size.

For supporting concepts, review Descriptive Statistics, Mean Median and Mode, Standard Deviation, Variance, Standard Error, Confidence Interval, Five Number Summary, Box Plot Interpretation and Histogram Interpretation.

Dunnett Comparison Decision Summary

The Dunnett comparison table focuses only on treatment-control comparisons. It does not include all possible group pairs because that is not the purpose of Dunnett’s Test.

Assumptions Before Dunnett’s Test

Dunnett’s Test is usually interpreted after a one-way ANOVA. Therefore, the basic ANOVA assumptions still matter. The outcome should be numeric, the groups should be independent, and the ANOVA error term should be reasonable for comparing treatment groups with the control.

For assumption support, use ANOVA Assumptions, Levene Test, Bartlett’s Test, Brown-Forsythe Test, Hartley F Max Test, Cochran C Test, Q-Q Plot Normality Check, P-P Plot Normality Check, Shapiro-Wilk Test and Outlier Detection.

SPSS Output Interpretation for Dunnett’s Test

The SPSS output for Dunnett’s Test should be read in this order: first the descriptive means, then the ANOVA table, then the Dunnett multiple-comparisons table. In the Dunnett table, the key columns are the mean difference, standard error, adjusted significance value and adjusted confidence interval.

SPSS Reading Order

SPSS Dunnett Interpretation

Python Chart-by-Chart Interpretation

The Python chart sequence explains Dunnett’s Test through group means versus control, adjusted confidence intervals, adjusted p-value decisions, distribution boxplots and ANOVA effect-size context.

Python Chart 1: Group Means vs Control

The group mean chart shows the control mean as a horizontal reference line at studytime group 1. Groups 2, 3 and 4 all stand above this control line. This visual immediately supports the main Dunnett question: whether each treatment group differs from the selected control.

The largest visible difference is group 3 versus control, followed by group 4 versus control and group 2 versus control. Because Dunnett’s Test is control-focused, the chart should be read vertically against the control line rather than as an all-pairs post hoc comparison.

Python Chart 2: Dunnett Mean Differences with Adjusted Confidence Intervals

The mean-difference chart shows the estimated differences between each comparison group and the control group. The vertical zero line is the no-difference reference. All three confidence intervals remain to the right of zero, which means all three comparison groups are significantly higher than the control group.

Group 3 has the largest treatment-control difference, around 2.38 grade points. Group 4 is also clearly above control, around 2.21 grade points. Group 2 is smaller but still clearly above control, around 1.25 grade points.

Python Chart 3: Dunnett Adjusted p-value Decision

The adjusted p-value chart confirms the confidence-interval result. Every treatment-control p-value is far below α = .05. Group 3 versus control has the smallest adjusted p-value, followed by group 2 versus control and group 4 versus control.

This chart is useful for reporting because it shows that the result is not borderline. The treatment-control differences remain statistically significant even after the Dunnett multiple-comparison adjustment.

Python Chart 4: Distribution by Group Boxplots

The boxplots show the spread and center of G3 inside each studytime group. The control group has a lower central pattern, while groups 3 and 4 are visibly higher. Group 2 also sits above the control group, though with overlap in the distribution.

This visual matters because Dunnett’s Test compares means, but the distribution tells whether the groups also show meaningful spread and overlap. The result supports a control-group difference, while the boxplot reminds the reader that individual scores still vary within every group.

Python Chart 5: ANOVA Effect Size Context

The effect-size chart gives the overall ANOVA context behind the Dunnett comparisons. Eta squared is about 0.0688, and omega squared is about 0.0643. These values indicate that studytime explains a meaningful but not overwhelming portion of G3 variation.

This is important for reporting because a post hoc result can be statistically significant without being practically large. The effect-size chart helps balance the conclusion: all treatment groups are significantly higher than the control, and the overall group effect is moderate enough to be educationally meaningful but not the only factor explaining final grade.

R Chart-by-Chart Validation

The R chart sequence validates the same Dunnett’s Test conclusion in a second software workflow. The R images use a colorful style, but the interpretation remains the same: groups 2, 3 and 4 are compared only with the control group, and all three comparison groups are significantly higher than control.

R Chart 1: Colorful Group Means vs Control

The R group mean chart confirms the Python result. The dashed control line marks studytime group 1, and groups 2, 3 and 4 all appear above that line.

This validates the main visual conclusion: the comparison groups have higher G3 means than the selected control group.

R Chart 2: Colorful Dunnett Mean Differences with Adjusted CIs

The R confidence-interval chart confirms that all treatment-control intervals remain to the right of zero. This means the differences are positive and statistically significant after Dunnett adjustment.

The interval for group 4 is wider than the others because group 4 has the smallest sample size. Even so, its interval remains above zero, so the group 4 versus control comparison is still significant.

R Chart 3: Colorful Dunnett Adjusted p-values

The R adjusted p-value chart confirms that the treatment-control comparisons are significant. The p-values are visibly far below the .05 threshold.

The chart supports the same final statement as Python: groups 2, 3 and 4 each differ significantly from control group 1.

R Chart 4: Colorful Distribution by Group

The R boxplot confirms the same distribution pattern. The control group has a lower central score than groups 3 and 4, while group 2 also trends higher than the control.

This distribution context supports the treatment-control comparison but also shows overlap among individual scores. That is why both visual and adjusted statistical results are needed.

R Chart 5: Group Size and Mean

The group-size chart shows why precision differs across comparisons. Group 2 has the largest sample size, group 1 is also large, group 3 is smaller, and group 4 is the smallest group.

The mean line rises sharply from group 1 to group 3 and stays high at group 4. This chart helps explain both the significant result and the wider uncertainty around group 4.

SPSS, R, Python and Excel Workflows for Dunnett’s Test

The same Dunnett’s Test workflow can be reproduced in SPSS, R, Python and Excel. SPSS can request Dunnett directly from the One-Way ANOVA post hoc menu. R can use multiple-comparison tools to compare treatments with a selected control. Python can calculate the ANOVA context and then use multiple-comparison logic for control-focused comparisons. Excel can support the mean-difference workflow, but SPSS or R is better for formal adjusted Dunnett p-values.

SPSS Workflow

R Workflow

Python Workflow

Excel Workflow

Code Blocks for Dunnett’s Test

SPSS Syntax for Dunnett’s Test

* Dunnett's Test in SPSS.
* Dependent variable: G3.
* Factor: studytime.
* Control group: studytime group 1.

TITLE "Dunnett's Test: G3 by Studytime with Group 1 as Control".

ONEWAY G3 BY studytime
  /STATISTICS DESCRIPTIVES HOMOGENEITY
  /POSTHOC = DUNNETT ALPHA(.05)
  /MISSING ANALYSIS.

EXAMINE VARIABLES=G3 BY studytime
  /PLOT BOXPLOT
  /COMPARE GROUPS
  /STATISTICS DESCRIPTIVES
  /CINTERVAL 95
  /MISSING LISTWISE
  /NOTOTAL.

OUTPUT EXPORT
  /CONTENTS EXPORT=VISIBLE
  /PDF DOCUMENTFILE="Dunnetts-Test-SPSS-Output.pdf".

Python Code for Dunnett-Style Control Comparisons

import pandas as pd
import numpy as np
from scipy import stats
from statsmodels.formula.api import ols
from statsmodels.stats.anova import anova_lm

df = pd.read_csv("dataset.csv")

df["G3"] = pd.to_numeric(df["G3"], errors="coerce")
df["studytime"] = df["studytime"].astype("category")

data = df[["G3", "studytime"]].dropna().copy()

# ANOVA context
model = ols("G3 ~ C(studytime)", data=data).fit()
anova_table = anova_lm(model, typ=2)

mse = anova_table.loc["Residual", "sum_sq"] / anova_table.loc["Residual", "df"]
df_error = anova_table.loc["Residual", "df"]

summary = data.groupby("studytime", observed=True)["G3"].agg(
    n="count",
    mean="mean",
    sd="std"
).reset_index()

print(summary)
print(anova_table)

control = "1"
control_mean = summary.loc[summary["studytime"].astype(str) == control, "mean"].iloc[0]
control_n = summary.loc[summary["studytime"].astype(str) == control, "n"].iloc[0]

rows = []

for _, row in summary.iterrows():
    group = str(row["studytime"])

    if group == control:
        continue

    treatment_mean = row["mean"]
    treatment_n = row["n"]

    diff = treatment_mean - control_mean
    se = np.sqrt(mse * (1 / treatment_n + 1 / control_n))
    t_value = diff / se

    # This raw t reference is shown for transparent calculation.
    # For final Dunnett adjusted p-values, validate against SPSS or R multiple-comparison output.
    raw_p = 2 * (1 - stats.t.cdf(abs(t_value), df_error))

    rows.append({
        "comparison": f"{group} vs control {control}",
        "control_mean": control_mean,
        "treatment_mean": treatment_mean,
        "mean_difference": diff,
        "standard_error": se,
        "t_value_reference": t_value,
        "raw_p_reference": raw_p
    })

dunnett_table = pd.DataFrame(rows)
print(dunnett_table)

# Verified adjusted p-value pattern from the report/charts:
# group 2 vs control: 2.794e-05
# group 3 vs control: 2.422e-09
# group 4 vs control: 0.0003436

R Code for Dunnett’s Test

# Dunnett's Test in R

library(tidyverse)
library(car)
library(multcomp)

df <- read.csv("dataset.csv")

df$G3 <- as.numeric(df$G3)
df$studytime <- as.factor(df$studytime)

data <- df %>%
  select(G3, studytime) %>%
  drop_na()

# Set studytime group 1 as control
data$studytime <- relevel(data$studytime, ref = "1")

# Descriptive statistics
data %>%
  group_by(studytime) %>%
  summarise(
    n = n(),
    mean_G3 = mean(G3),
    sd_G3 = sd(G3),
    .groups = "drop"
  )

# ANOVA model
model <- aov(G3 ~ studytime, data = data)
summary(model)

# Assumption context
leveneTest(G3 ~ studytime, data = data)

# Dunnett's Test: compare each group with control group 1
dunnett_result <- glht(model, linfct = mcp(studytime = "Dunnett"))

summary(dunnett_result)
confint(dunnett_result)

# Optional compact letters are not the main Dunnett output.
# Dunnett focuses on treatment-control comparisons.

Excel Notes for Dunnett’s Test

Excel support workflow:

1. Arrange the data:
   G3 | studytime

2. Create group summaries:
   n, mean, standard deviation, standard error.

3. Select the control group:
   studytime group 1

4. Calculate treatment-control differences:
   Group 2 - Group 1
   Group 3 - Group 1
   Group 4 - Group 1

5. Calculate the standard error:
   =SQRT(MSE*(1/n_treatment+1/n_control))

6. Calculate the t statistic:
   =mean_difference/standard_error

7. Use validated Dunnett adjusted p-values or Dunnett critical values:
   Group 2 vs Control: significant
   Group 3 vs Control: significant
   Group 4 vs Control: significant

8. Report:
   Each comparison group has significantly higher mean G3 than control group 1.

APA Reporting Wording

When reporting Dunnett’s Test, first state the control group, then report the treatment-control comparisons. Do not report it as an all-pairs post hoc test.

Common Mistakes

When to Use Dunnett’s Test

Use Dunnett’s Test when your design has one control group and several treatment or comparison groups. It is especially useful in experiments, clinical studies, education studies and treatment evaluation designs where the main question is whether each treatment performs differently from a baseline condition.

Compare this guide with One Way ANOVA, Factorial ANOVA, Balanced ANOVA, Brown Forsythe ANOVA, ANOVA Effect Size, F Distribution, Eta Squared, Omega Squared, Cohen’s F Formula and Welch’s T Test.

Downloads and Resources for Dunnett’s Test

Use these resources to reproduce the Dunnett’s Test workflow. The Python report, R report and SPSS output PDF are included as verification files. Script and workbook placeholders can be replaced after the final downloadable files are uploaded to the WordPress Media Library.

Related Guides