Advanced ANOVA Techniques

How to Perform Advanced ANOVA Techniques Using DataStatPro: Repeated Measures and Mixed Designs

Learning Objectives

By the end of this tutorial, you will be able to:

Understand when to use repeated measures ANOVA vs mixed designs
Set up and conduct repeated measures ANOVA in DataStatPro
Perform mixed-design ANOVA with between and within-subject factors
Interpret sphericity assumptions and apply corrections
Report results in publication-ready format

When to Use Advanced ANOVA Techniques

Repeated Measures ANOVA

Use repeated measures ANOVA when:

Same participants measured multiple times (longitudinal studies)
Multiple conditions tested on same subjects (within-subject designs)
Pre-post-follow-up measurements
Time series with multiple measurement points

Mixed-Design ANOVA

Use mixed-design ANOVA when:

Combining between-subject and within-subject factors
Different groups measured over time
Treatment groups with repeated assessments
Factorial designs with repeated measures on one factor

Step-by-Step Guide: Repeated Measures ANOVA

Step 1: Data Preparation

Navigate to ANOVA Analysis
- Go to Inference → ANOVA Tests
- Select Repeated Measures ANOVA
Prepare Your Data
- Ensure data is in wide format (one row per participant)
- Each time point should be a separate column
- Include participant ID column

Step 2: Variable Selection

Select Repeated Measures Variables
- Choose all time point columns (e.g., Time1, Time2, Time3)
- Ensure variables are numeric and properly labeled
Optional: Add Between-Subject Factors
- Select grouping variables if conducting mixed design
- Choose categorical variables for group comparisons

Step 3: Assumption Checking

Sphericity Assessment
- Review Mauchly's test of sphericity
- Check epsilon values (Greenhouse-Geisser, Huynh-Feldt)
- Examine sphericity assumption violations
Normality and Outliers
- Check residual normality plots
- Identify potential outliers
- Assess homogeneity of variance

Step 4: Running the Analysis

Execute Analysis
- Click Run Analysis
- Review output tables and plots
- Check for convergence and warnings
Interpret Main Results
- Examine within-subjects effects table
- Review between-subjects effects (if applicable)
- Check effect sizes (partial eta-squared)

Step-by-Step Guide: Mixed-Design ANOVA

Step 1: Design Setup

Access Mixed Design Options
- Select Mixed-Design ANOVA from ANOVA menu
- Configure between and within-subject factors
Factor Configuration
- Between-Subject Factor: Group membership variable
- Within-Subject Factor: Repeated measures variables
- Specify factor levels and labels

Step 2: Model Specification

Main Effects and Interactions
- Include main effect of between-subject factor
- Include main effect of within-subject factor
- Specify interaction between factors
Contrast Planning
- Set up planned comparisons
- Choose contrast coding (treatment, sum, etc.)
- Specify custom contrasts if needed

Step 3: Post-Hoc Analysis

Multiple Comparisons
- Apply Bonferroni correction for multiple tests
- Use Tukey HSD for pairwise comparisons
- Consider false discovery rate (FDR) control
Simple Effects Analysis
- Test group differences at each time point
- Examine time effects within each group
- Interpret interaction patterns

Real-World Example: Clinical Trial Analysis

Scenario

A clinical trial comparing two treatments (Drug A vs Placebo) with measurements at baseline, 4 weeks, 8 weeks, and 12 weeks.

Data Structure

Participant | Group   | Baseline | Week4 | Week8 | Week12
001         | Drug_A  | 45       | 42    | 38    | 35
002         | Placebo | 44       | 43    | 42    | 41
...

Analysis Steps

Mixed-Design Setup
- Between-subject factor: Group (Drug_A, Placebo)
- Within-subject factor: Time (Baseline, Week4, Week8, Week12)
Key Results Interpretation
- Main effect of Time: Overall change across time points
- Main effect of Group: Overall difference between treatments
- Group × Time interaction: Different change patterns by treatment
Follow-up Analyses
- Simple effects: Group differences at each time point
- Trend analysis: Linear, quadratic change patterns
- Effect size calculation: Clinical significance assessment

Interpreting Results

Understanding F-Statistics

Within-Subjects Effects: Tests changes over time
Between-Subjects Effects: Tests group differences
Interaction Effects: Tests if group differences vary over time

Effect Size Interpretation

Partial Eta-Squared (ηp²):
- Small: 0.01
- Medium: 0.06
- Large: 0.14

Sphericity Corrections

Greenhouse-Geisser: Conservative correction for sphericity violations
Huynh-Feldt: Less conservative, use when epsilon > 0.75
Lower-bound: Most conservative, use for severe violations

Publication-Ready Reporting

Results Section Template

"A 2 × 4 mixed-design ANOVA was conducted with Group (Drug A, Placebo) as the between-subjects factor and Time (Baseline, Week 4, Week 8, Week 12) as the within-subjects factor. Mauchly's test indicated that the assumption of sphericity was violated (χ² = 12.45, p = .006), therefore Greenhouse-Geisser corrected degrees of freedom were used (ε = 0.78).

There was a significant main effect of Time, F(2.34, 140.4) = 15.67, p < .001, ηp² = .21, indicating overall improvement across time points. The main effect of Group was also significant, F(1, 60) = 8.92, p = .004, ηp² = .13, with Drug A showing greater improvement than Placebo.

Most importantly, there was a significant Group × Time interaction, F(2.34, 140.4) = 4.58, p = .008, ηp² = .07, indicating that the treatment groups showed different patterns of change over time."

APA Style Table

Table 1
Mixed-Design ANOVA Results for Treatment Efficacy

Source                    df      F       p       ηp²
Between-subjects
  Group                   1      8.92    .004    .13
  Error                   60

Within-subjects
  Time                    2.34   15.67   <.001   .21
  Time × Group            2.34   4.58    .008    .07
  Error(Time)             140.4

Troubleshooting Common Issues

Problem: Sphericity Violation

Solution: Apply appropriate correction (Greenhouse-Geisser or Huynh-Feldt)

Problem: Missing Data

Solution: Use mixed-effects models or multiple imputation

Problem: Unbalanced Design

Solution: Consider Type III sum of squares or mixed-effects approach

Problem: Non-normal Residuals

Solution: Transform data or use robust ANOVA methods

Frequently Asked Questions

Q: When should I use repeated measures vs mixed-effects models?

A: Use repeated measures ANOVA for balanced designs with complete data. Use mixed-effects models for unbalanced designs, missing data, or complex correlation structures.

Q: How do I handle missing data in repeated measures?

A: Options include listwise deletion (complete cases only), multiple imputation, or switching to mixed-effects models which handle missing data naturally.

Q: What if sphericity is severely violated?

A: Consider multivariate ANOVA (MANOVA) approach or use mixed-effects models which don't assume sphericity.

Q: How many participants do I need?

A: Minimum 10-15 per group, but power analysis should guide sample size based on expected effect size and desired power.

Q: Can I include covariates in repeated measures ANOVA?

A: Yes, use ANCOVA with repeated measures, but ensure covariates don't change over time or model time-varying covariates appropriately.

Next Steps

After mastering advanced ANOVA techniques, consider exploring:

Mixed-effects models for complex longitudinal data
Multivariate ANOVA (MANOVA) for multiple outcomes
Structural equation modeling for complex relationships
Bayesian ANOVA approaches for uncertainty quantification

This tutorial is part of DataStatPro's comprehensive statistical analysis guide. For more advanced techniques and personalized support, explore our Pro features.