Experimental Design Notes
Introduction to Experimental Design
Importance of Experimental Design
Enables an organized and efficient plan for a project and its experiments.
Allows understanding of the rationale and purpose of the experiments.
Promotes high-quality research that is reliable and replicable.
Fundamentals Before Experiments
Select a discipline, area, and topic, and define its scope.
Conduct a literature search/review of the topic and scope.
Identify the research issue/problem to address, elaborating on it and avoiding focus on knowledge gaps as mere academic curiosity.
Formulate a research question linked to the issue/problem.
Construct a hypothesis to address the research issue/problem and questions.
Propose an explanation for a phenomenon or problem.
Ensure testable predictions.
Determine how it can be tested.
The hypothesis is either accepted or rejected based on collected data.
Direction:
Hypothesis → data
Hypothesis ← data
Elaborate on significance: Why does addressing this matter? To whom is this important?
Field
Discipline
Knowledge advancement (basic, foundational, apply)
Stakeholders
Industry
Formulate and narrow the aims of the study, ensuring they link to the central question.
Defining Rationale and Scope
Define the rationale and overall details of the research approach.
Define the scope—specific and narrow the topic.
Assess the feasibility and resolution of the approach to address the study aims.
Draft a proposal and seek feedback from colleagues/research team.
Designing and Planning Experiments
Define parameters for the experiment by being clear and concise in wording.
Clearly defining terms helps focus on experimental methods and avoid ambiguity.
Ensures more accurate results and less flexibility in experimental design, increasing accuracy.
Measurements
Decide whether to perform a mensurative or manipulative experiment.
Mensurative experiment: involves making measurements at different times or in different areas.
Manipulative experiment: involves physically altering a treatment group, and thus always has two or more treatments.
Sample Size
Choose an appropriate sample size fitting for the results.
Smaller sample sizes produce inaccurate results for generalization.
A smaller sample size produces a smaller effect size measure, which is the efficacy of the treatment.
Control Groups
Introduce a control group.
In biology, systems exhibit temporal change, a potential influencing third variable.
A control is necessary to isolate changes to the experimental treatment alone.
Randomize Assignment
Randomizing sample units to different treatment groups avoids experimenter bias.
Randomization is critical as it intersperses the samples being tested.
Replicates
The number of replicates varies with design but ensures precision in experiments.
Sample Distribution
Ensure samples are dispersed in space or time to avoid pseudoreplication.
This ensures replicates are statistically independent.
Avoid quantifying samples from the same unit as independent, as it is not genuine replication.
Statistics
Use linear regression-based analysis first before introducing analysis of variance (ANOVA).
Linear regression is more powerful to analyze obtained data and indicates how dependent variables change with the independent variable.
Refrain from deducing results based on P value alone.
The p-value indicates the confidence interval in statistics, but it does not indicate how a system actually changes.
Effect size measures are more meaningful in ecology and should be given more weight in findings.
Experimental Design in Conservation Genetics/Genomics
What is the topic and scope of your research questions?
Summary of Experimental Design
Define the problem
Define the hypothesis
Objectives
Outline the overall approach
Obtain feedback
Design experiments
Conduct experiments
Analyze data
Interpret results
Research Proposal Structure
Summary/Abstract: Clear and concise summary of the research project, aligning with the body of the report and accurately portraying the findings within a broader conservation context.
Introduction: Research question placed within the context of the conservation challenge, issue/problems, and the broader literature.
Hypothesis (hypotheses): Appropriately worded hypothesis relevant to the broader research question and answerable by the data analysis approach.
Methods: Clear summary of the data collection/generation approach. Data analysis approach recounted in sufficient detail to allow a third party to repeat the analysis.
Results & Data Presentation: The most relevant data analysis approach given the available data and the research question/hypothesis. Results presented in appropriate format, including well-labeled diagram/graphs, with necessary measures of data variability.
Discussion and management recommendations: Research findings placed within the context of the broader literature; appropriate management/future research recommendations made, aligning with the results.
References: Appropriate evidence base to support all sections of the text; references well formatted (in-text and reference list).
Group work: Teamwork, setting clear/relevant milestones, appropriate and inclusive communication strategy, response to advisor feedback, meeting project milestones as outlined in group work agreement.
Recommended Readings
Papers on randomization and balancing of sampling and molecular ecology experiments.
Sections 1.1 Experimental Design Background and 1.2 Sampling Design of An Introduction to Statistical Analysis in Research: With Applications in the Biological and Life Sciences.
Bálint, M, Márton, O, Schatz, M, Düring, R‐A, Grossart, H‐P. Proper experimental design requires randomization/balancing of molecular ecology experiments. Ecol Evol. 2018; 8: 1786– 1793. https://doi.org/10.1002/ece3.3687