Experimental Design and Data Analysis

Pairwise Comparisons: Witchweed and Fertilizer Interaction

  • Using anova models to produce interaction plots to guide recommendations for fertilizer or weed control.
  • Tukey's tests can be performed to further analyze combinations.
  • P-values are adjusted to account for family-wise error rates due to multiple comparisons.
  • Factorial designs are powerful as they address multiple research questions, reduce type one errors, and provide a better understanding of mechanisms.

Understanding Main Effects and Interactions

  • Grasp the concept of main effects and interactions.
  • Learn to interpret ANOVA tables by starting from the highest order interaction term; if it's not significant, move up to the next higher orders or main effects.
  • Avoid just listing significant findings from an ANOVA table when a significant interaction is present, focus on what's truly important.

Checking Model Assumptions with Residuals

  • Always check model assumptions for ANOVA.
  • Use residuals from models with graphical approaches like residuals versus fitted plots and QQ plots.
  • Residuals vs Fitted plots: testing equal variance assumption.
  • QQ plots: testing normality assumption.
  • Residuals are essential for testing assumptions; using raw data will result in deductions.

Interpreting Confidence Intervals in E-Means Plots

  • Confidence intervals in e-means plots help visualize differences.
  • Overlapping confidence intervals indicate non-significance.

Addressing Assumption Violations in ANOVA

  • Real-world data often violates normality assumptions.
  • ANOVA is robust against departures from normality, especially in balanced designs (equal replication across treatment levels).
  • Equal variance assumption is more critical than normality assumption.
  • If transformations (like log transformation for count data or arcsine for proportion data) fail to improve assumptions, acknowledge the limitation and interpret results cautiously.

Transformations in Data Analysis

  • Try a log transformation for count data.
  • Use a square root transformation for a harsher effect.
  • Use arcsine transformation for proportion data.
  • Avoid excessive transformations to maintain data interpretability.
  • If assumptions fail, acknowledge it and proceed with caution.
  • Cite papers supporting the robustness of ANOVA against deviations from normality.

Kruskal-Wallis Test

  • Kruskal-Wallis test, is an alternative for simple one-way ANOVA designs.

Second Project Introduction

  • The second project involves using real-world data from a published paper to conduct a full scientific paper analysis.
  • Download the pre-processed data from Canvas, not directly from the paper.
  • Choose one of the provided data sheets or create your own data set.
  • Formulate a scientific question and conduct a complete analysis.

Scientific Questions and Experimental Design

  • Develop a question, determine the experimental design, treatment design, and model equation.
  • Use ANOVA and multiple regression techniques to answer the questions.
  • Design questions that can be answered using these techniques.
  • Focus on residuals, not raw data.

Scientific Writing Guide

  • Follow a set formula for writing scientific papers.
  • Refer to first-year notes on scientific writing.

Data Set Metadata

  • Use the metadata sheet in the Excel file to understand the data set variables and measurements.
  • Metadata helps ensure proper data usage and interpretation when using datasets collected by others.

Template and Formatting

  • Use the provided Quattro or R Markdown template for the report.
  • Ensure "embed resources" is set to true in Quattro to render plots.

Utilizing AI Tools

  • Use AI tools cautiously for coding and plot improvements.
  • Avoid using AI for interpretation or analysis.
  • Acknowledge AI usage in the acknowledgement section, including prompts and verification methods.

Referencing and Literature

  • Format references consistently, following a specified style.
  • Cite primary literature (journal articles, textbooks).
  • Avoid web pages as primary sources unless complementing primary sources.

Appendix

  • Use the appendix for supplementary information, such as assumption testing plots.
  • Label figures in the appendix as "Figure A1", "Figure A2", etc.

Model Assumptions Testing Location

  • Test model assumptions in the methods section.
  • Report whether assumptions were met or violated, and reference appendix plots.

Post-Hoc Tests

  • Post-hoc tests can be conducted after the ANOVA.