RSM 3/12/26

Discussion about examples: pairwise contrast and multiple range.
Introduction of the concept of contrast with an example involving fabric treatments.

Five treatments of fabrics with varying cotton percentages:
- 15% cotton
- 20% cotton
- 25% cotton
- 30% cotton
- 35% cotton
A total of five samples per treatment is used to measure the strength of the fabrics.
The total number of observations is calculated as 25.

Definition of degrees of freedom for the treatments:
- Treatments: 4 (as the number of treatments - 1)
- Total: 24 (total samples - 1)
Partitioning of the ANOVA data:
- Treatment sum of squares = 475.76
- Error sum of squares = 161.2
Calculation of mean squares:
- Mean square for treatments = $\frac{475.76}{4} = 118.94$
- Mean square for error = $\frac{161.2}{24} = 8.06$
F-value calculation for ANOVA:
- F-value = $\frac{118.94}{8.06} = 14.76$

Definition of contrasts and their coefficients explained:
Four hypotheses for testing contrasts set up:
1. Hypothesis regarding the strength of 30% cotton and 35% cotton being the same as that of 15% and 25% cotton.
2. Does 15% have the same strength as 25%?
3. Is 20% cotton strength the same as the average strength of 15%, 25%, 30%, and 35% cotton?
4. Additional contrast setups and specific configurations of coefficients analyzed.
Orthogonality defined:
- A set of contrasts is orthogonal if the coefficients add up to zero.

Example of how to set up coefficients for contrasts:
- Coefficients explained step by step for each of the hypotheses.
For contrast hypotheses definitions, the significance of the positive and negative values discussed.

Clarification of how to calculate sums:
- Sum of values for each treatment provided:
- For treatments: 1: 49, 2: 88, 4: 108, 5: 54
Calculation of sums multiplied by their respective coefficients to yield results for each contrast:
- Specified breakdown of calculations, including detailed coefficient applications and responses per hypothesis.

Explanation of contrasts squared calculations:
- For contrast one, calculation involves squaring the differences and dividing by calculated coefficients to yield sum of squares.
- Similar calculations outlined for contrast two and three, with a breakdown of the mathematical process.
Specific numerical examples for squares and resulting calculations itemized:
- Example calculations leading to specific numerical conclusions.

Evolution into the ANOVA table including each contrast and their mean sum of squares calculations.
- Detail how to derive the F-value from mean squares from each hypothesis.
Review of relevant statistical properties like p-values derived from contrasts.

Outcomes discussed from contrasts and their implications:
- P-values less than 0.05 indicate significance of differences between treatments.
- Summary of findings across each hypothesis in relation to fabrics strengths.
Suggestions for reporting the output to imply clear distinctions between means derived from contrasts, emphasizing best practices for analysis.
Addressing potential confusion in reporting statistical outcomes and methods for clarity.

Discussion on linear vs. quadratic effects and importance of understanding their characteristics in regards to increased cotton content.
- Detailed analysis of coefficients for linear and quadratic regression outlined, noting the distances and ranges between variable levels in the context of cotton's effects on fabric strength.
Implications of higher degrees (cubic effects) in future studies highlighted to ascertain patterns and their meaning.

Importance of defining coefficients in computational settings (SAS) emphasized as critical for accurate analyses.
Coefficients clarified again, stressing the need for careful declaration of zeros when dealing with real-world application datasets.

Recognition of the significance of effective contrast methods in statistical analysis and study reporting.
Encouragement to retain key concepts discussed while acknowledging the potential complexity of some advanced topics.
Note on designing a framework (like LSD or pairwise compression techniques) to correctly interpret and process multiple range comparisons without losing accuracy or insight into significant findings.