4/7/25 psych Factorial ANOVA & Post Hoc Testing in Research Analysis
Core Group Recognition
Acknowledgment of students' commitment
Importance of attendance in late semester classes
Introduction to Factorial ANOVA
The lesson begins with a focus on factorial ANOVA, a key concept for the course's final module (Module 10).
Connection to Previous Material:
Simple linear regression relates to ANOVA just as multiple regression relates to factorial ANOVA.
Factorial ANOVA considers multiple independent variables (factors) simultaneously.
Design Research Studies
Importance of research design in selecting appropriate analysis types.
Noted as a significant aspect of understanding ANOVA.
Emphasis on definitions and primary concepts today, laying groundwork for upcoming evaluations.
F Distribution
Difference between F distribution and other distributions (e.g., Z, T).
F distribution cannot be negative; it starts at 0 and progresses to positive infinity.
Key point: When comparing variances, a larger numerator and a smaller denominator results in a significant F ratio, indicating group differences.
Test Statistics Overview
Types of Test Statistics:
Z Test: Used when population standard deviation is known for comparing one sample. Rarely applied.
T Test: Appropriate when population information is unknown. Useful for comparing sample differences.
Fisher's Ratio (ANOVA): Appropriate when comparing means among three or more groups.
Post Hoc Tests
Definition: Conducted after ANOVA to determine which specific means differ (Latin for 'after the fact').
Necessary because ANOVA indicates at least one mean is different but does not specify which.
Analogy: The F test is like a signal of a problem requiring further investigation, the post hoc tests then explore the problem further.
Importance and Limitations of Post Hoc Tests
Advantages:
Save time by only testing when ANOVA shows significant differences.
Reduces the number of analyses needed compared to multiple independent tests.
Challenges:
Increased risk of false positives if multiple tests are conducted.
Some variables might be overlooked in planned contrasts, whereas exploratory post hoc may yield unexpected results.
Subtypes of Post Hoc Tests
Different approaches exist in performing post hoc tests, each varying in stringency:
Fisher's LSD (Least Significant Difference): Conducted only if F is significant; it seeks the minimal significant difference.
Tukey’s HSD (Honest Significant Difference): More conservative than Fisher’s LSD and controls for type I error risk across multiple comparisons.
Bonferroni correction: Adjusts the alpha level depending on the number of tests being conducted to minimize false positives.
Comparative Analysis Between Techniques
Both planned contrasts and exploratory post hoc tests have strengths and weaknesses.
Illustration of Differences: Analogous to exploring health symptoms, a precise diagnosis is essential for effective treatment.
Highlighted that correlational studies or exploratory findings are subject to interpretation bias, similar to medical diagnoses based on symptom interpretation.
Implication of Variances in ANOVA
ANOVA works by splitting variances into between-group and within-group components.
It reflects group differences relative to individual variances, allowing better analysis of interactions between variables.
Factorial Designs
Discussed the necessity of factorial designs when several factors influence an outcome.
Definition Clarity:
Factors (independent variables) can have multiple levels and interactions that change how we analyze group behavior.
Encouragement to think of everyday examples of crossed factorial designs, like combinations of physical traits (e.g., eye color and hair color).
Summary:
If dealing with multiple means (>2), prefer ANOVA over t-tests, although for 2 groups, they are mathematically interchangeable.
Emphasis on understanding group behaviors through the lens of variance and the necessity of follow-up tests to clarify group relationships.
This foundational knowledge sets the stage for more complex analyses and deeper understanding of statistical relationships in research.
Core Group Recognition
Detailed list of names to acknowledge each student's commitment.
Specifics on attendance requirements, including potential penalties for absences.
Additional support provided to students with attendance issues.
Introduction to Factorial ANOVA
Comprehensive review of factorial ANOVA, detailing its applications, assumptions, and interpretations.
Elaboration on the relationship between simple linear regression and ANOVA, as well as multiple regression and factorial ANOVA.
Explanation of how factorial ANOVA can handle multiple independent variables and interactions simultaneously; use examples.
Design Research Studies
Detailed discussion on how research design impacts the choice of statistical analysis.
Overview of different experimental designs (e.g., randomized control trials, cohort studies) and their relevance to ANOVA.
Discussion on definitions and primary concepts, setting up examples and mini-quizzes for upcoming evaluations.
F Distribution
In-depth comparison of the F distribution with Z and T distributions, highlighting key differences and similarities.
Explanation of why the F distribution starts at 0 and extends to positive infinity, including the concept of non-negativity.
Illustration on how variances affect the F ratio, emphasizing the importance of large numerators and small denominators in detecting group differences.
Test Statistics Overview
Types of Test Statistics:
Z Test: Expanded discussion on cases where population standard deviation is known, including real-world examples.
T Test: Further explanation on using the T test when population information is unavailable and its applications in comparing sample differences.
Fisher's Ratio (ANOVA): Detailed guide on applying ANOVA to compare means across three or more groups, including assumptions and limitations.
Post Hoc Tests
Clearer definition of post hoc tests and why they are necessary after ANOVA to pinpoint specific mean differences.
Use of accessible language to explain that ANOVA only signals the existence of differences without specifying which groups differ.
Extension of the analogy to real-world scenarios.
Importance and Limitations of Post Hoc Tests
Advantages:
Further discussion on how post hoc tests save time by focusing on significant differences identified by ANOVA.
Explanation of how post hoc tests reduce the overall number of analyses needed, lowering the chance of error.
Challenges:
Elaborate risk of false positives with multiple tests, including methods to mitigate such risks (e.g., Bonferroni correction).
Detailed exploration of scenarios where planned contrasts might overlook important variables compared to exploratory post hoc tests.
Subtypes of Post Hoc Tests
Detailed comparison of different post hoc test approaches and their stringency.
Fisher's LSD (Least Significant Difference): Provide conditions under which Fisher’s LSD is suitable and its limitations.
Tukey’s HSD (Honest Significant Difference): Clarify how Tukey’s HSD provides better control over type I error across multiple comparisons.
Bonferroni correction: Further explain how to adjust the alpha level based on the number of tests to better manage false positives.
Comparative Analysis Between Techniques
Detailed analysis of the strengths and weaknesses of both planned contrasts and exploratory post hoc tests.
Enhanced analogy about health symptoms to emphasize the need for precise diagnosis for effective treatment.
Discussion on potential interpretation biases in correlational studies or exploratory findings, paralleling medical diagnoses.
Implication of Variances in ANOVA
Comprehensive explanation of how ANOVA divides variances into between-group and within-group components.
Discussion on how these components help in analyzing interactions between variables by reflecting group differences relative to individual variances.
Factorial Designs
Strengthen reasons for using factorial designs when multiple factors influence an outcome.
Definition Clarity:
Provide more examples of factors and levels, explaining how interactions modify the analysis of group behavior.
Engage learners by asking them to provide their own examples of crossed factorial designs.
**Summary