Experimental Design: Hypotheses, Variables, and Control
Developing a Testable Hypothesis and Experimental Design
The Foundation: Rationale for a Hypothesis (-1 Step)
Hypotheses must be grounded: Before stating a testable hypothesis (1^{st} step), a rational reason for conducting the experiment is essential. This prevents studies from being deemed pointless and rejected by reviewers or publishers.
Sources of Rationale:
Pure Observation: Everyday observations (e.g., observing people's behavior on the street) can lead to hypotheses.
Example: Laurier brings dogs to campus during final exams. Observation: Dogs are present during stressful periods. Inference: Laurier probably does this because it's good for students. Conclusion: Pets probably improve health.
Theory from Another Discipline: Existing theories can provide support.
Example: Pets serve as social support animals. Theory: Social support increases health (SS \rightarrow H). Conclusion: If pets provide social support, they probably improve health.
Past Research: Previous studies can inform new hypotheses.
Example: Research shows a correlation between pet ownership and happiness. Inference: This correlation might extend to health as well.
All these behind-the-scenes steps are crucial before formally stating a hypothesis.
Components of a Basic Hypothesis (1^{st} Step)
A basic hypothesis consists of:
Two Variables: The elements being studied.
A Relationship between those two variables: How one variable influences the other (e.g., increases, decreases, improves, reduces).
Importance of a Relationship: A hypothesis should not vaguely state "pets affect health." It must specify how they affect health (e.g., "pets improve health" or "pets decrease something related to health"). The word "effect" is generally too vague for a scientific hypothesis statement.
Understanding Variables: Independent (IV) and Dependent (DV)
Variable Definition: Anything that changes or can be changed/measured.
Independent Variable (IV):
Definition: The variable that is manipulated by you, the experimenter.
Role: The "cause" in a cause-and-effect relationship.
Example (Hypothesis: Pets improve health): "Pets" would be the IV, as the experimenter would create a situation involving pets (e.g., giving somebody a pet).
Mnemonic (MIX): Manipulated, Independent, X-axis (graphed on the x-axis).
Dependent Variable (DV):
Definition: The variable that is measured by you, the experimenter.
Role: The "effect" in a cause-and-effect relationship; it responds to changes in the IV.
Example (Hypothesis: Pets improve health): "Health" would be the DV, as the experimenter would measure people's health.
Mnemonic (DRY): Dependent, Responding/Result/Outcome, Y-axis (graphed on the y-axis).
Operational Definitions: Making Variables Concrete
Purpose: To define specific, concrete actions or operations that will be used to create your independent variable and measure your dependent variable.
Why they are crucial:
Clarity: Moves variables from vague, theoretical concepts (e.g., "health") to measurable realities.
Replication: Enables other researchers to replicate the study by providing a clear "recipe" for how it was conducted. Without replication, scientific findings lack validation and importance.
Process: Transforming vague/theoretical variables into concrete, measurable ones.
More precise definitions lead to better hypotheses and greater replicability.
Operational Defining the Dependent Variable (DV): "Health"
Initial Vague Hypothesis: "Pets improve health."
Problem: "Health" is subjective and can mean many things to different people (physical, mental, specific conditions, etc.).
Improvement 1: Adding "Feelings": "Pets improve feelings of health."
Pro: More specific measurements (e.g., asking participants on a scale from 1 to 5 "How healthy do you feel?"). This provides more information on how health is being measured.
Con: Still too vague. "Healthy" remains subjective to participants (e.g., physical vs. mental health). Participants need concrete definitions.
Improvement 2: Specifying Behaviors: "Pets decreased doctor visits and colds."
Pro: Highly concrete and specific. Participants report tangible behaviors (number of doctor visits, number of colds this semester), reducing subjectivity.
Pro: Incorporates multiple measures of health (doctor visits, colds). Measuring a dependent variable in different but related ways provides a more comprehensive and accurate assessment than a single question.
Con: Relies on self-report, which can be less reliable due to:
Forgetfulness (participants may not accurately recall past events).
Stigma (desire to appear healthier, leading to dishonest reporting).
Lack of knowledge (e.g., confusing a cold with the flu).
Application: This form of hypothesis is well-suited for survey-based studies, where direct manipulation of variables might be unethical or impractical.
Improvement 3: Objective, Evidence-Based Measures: "Pets improve blood pressure, hip-to-waist ratio, ECG results."
Pro: Maximum specificity and objectivity. These are biologically verifiable, evidence-based measures with decades of established links to health.
Con: Logistical challenges: Requires expensive equipment, trained personnel, and significant financial resources to measure hundreds of people. Not always accessible or affordable for every study.
Recap DV Operational Definition: The concrete, measurable definition of the outcome variable that is thought to be caused by the independent variable.
Operational Defining the Independent Variable (IV): "Pets"
Purpose: To define how the IV will be presented or created for the experimental and control groups.
Basic Experiment Structure: The most basic experiment has two groups:
Experimental Group: Receives the presence of the independent variable (the "treatment").
Control Group: Receives the absence of the independent variable (no "treatment," a placebo, or an alternative).
Initial Vague IV (from hypothesis "Pets improve blood pressure"): "Pets."
Problem: What constitutes a "pet"? A rock (as in the 1970s "pet rock")? An animal pet?
Improvement 1: Specifying Animal Type: "Animal pets."
Problem: What kind of animal pet? Dogs, fish, cats, snakes, ferrets, skunks? Different animals have different implications for care, interaction, etc.
Improvement 2: Specifying Species: "Dog."
Problem: What kind of dog? A Great Dane might have different exercise needs than a Chihuahua, or a Border Collie different from a Chihuahua. Specificity is required.
Improvement 3: Further Specificity: "Small adult dog."
Rationale: Puppies are typically very stressful (potentially raising blood pressure), and large dogs have intense exercise needs. A "small adult dog" balances logistics and reduces confounding stress factors.
Operational Definition for IV (Experimental Group): The presence of a small adult dog.
The Control Group: Ensuring "All Else Equal"
Problem with Simple "Absence of IV": If an experimental group receives a small adult dog (requiring walks twice a day, social interaction, financial costs of food, etc.) and a control group receives simply "no small adult dog," any observed difference in blood pressure might not be due to the dog alone. It could be due to increased exercise from walking the dog, reduced stress from social interaction, or financial considerations of pet care.
Goal of a Good Control Group Operational Definition: To ensure the control group is as similar as possible to the experimental group, except for the independent variable.
Strategy: Have the control group engage in activities that mimic the extraneous behaviors associated with the experimental group's interaction with the IV.
Example (Pets and Blood Pressure): Instead of simply "no small adult dog," give the control group something else that requires similar care, interaction, and associated behaviors but isn't a live animal.
Conceptual Example from Lecture: A "Furby" (a toy from the 90s requiring feeding, "walking," talking to, etc.).
Operational Definition for Control Group: Caring for a Furby.
Resulting New, Improved Hypothesis: "Will caring for a small adult dog lead to lower blood pressure than caring for a Furby?"
Benefits: Clearly defines both IVs (caring for a small adult dog vs. caring for a Furby), clarifies the DV (blood pressure), and, by design, implies that potentially confounding factors (like amount of activity, social interaction, financial responsibility of "caring for something") are equalized across groups.
Logical Conclusion: If all else is equal (activities, responsibilities are matched), and the experimental group still shows lower blood pressure, then the difference must be caused by the specific, independent variable (the type of companion - a dog vs. a Furby), not other associated behaviors.
Controlling Extraneous Variables: Random Assignment
Extraneous Variables: Any variables other than the IV that could potentially affect the DV. Control groups aim to equalize these through careful design.
Random Assignment Defined: All participants have an equal chance of being assigned to either the experimental condition or the control condition.
Purpose: To minimize the likelihood that pre-existing, uncontrolled differences between participants in different groups (e.g., medical conditions, lifestyle, income, diet) are what cause changes in the dependent variable, rather than the independent variable.
How it works: By giving everyone an equal chance of assignment, random assignment statistically ensures that important extraneous variables are distributed roughly equally across all groups.
Example: If some participants have a medical condition affecting blood pressure, random assignment should lead to a similar number of such participants in the experimental and control groups, thus equalizing this factor across conditions.
Consequences of Non-Random Assignment (Poor Design Example):
Scenario (Convenience Sampling): Experimental group recruited from "doggy daycare owners"; Control group recruited from "PrEP students who don't own dogs."
Experimental Group Characteristics: Likely higher income (can afford dog insurance, trainers, grass-fed beef, daycare), potentially lower stress, good blood pressure.
Control Group Characteristics: Likely lower income (maxed out credit card, only exercise is walking, survives on mac and cheese, works double shifts), potentially higher stress, high blood pressure.
Problem: Any observed difference in blood pressure could be explained by income/lifestyle (a significant extraneous variable) rather than pet ownership. This creates a confound, making it impossible to logically conclude that the IV caused the DV.
Benefits of Random Assignment:
Equalization: Extraneous variables (e.g., income, health conditions, lifestyle factors) are randomized across groups.
"All Else Equal" Achieved: The only systematic difference between groups is the presence/absence of the IV.
Causal Inference: Allows researchers to logically conclude that the independent variable caused the observed differences in the dependent variable.
Addressing Imperfect Randomization: If, despite random assignment, groups are not perfectly equal on a key extraneous variable, researchers can measure these variables (e.g., income, other health issues) and use statistical control to account for their influence in the analysis.
Participant Recruitment and Consent: During the recruitment stage, statements should be vague enough to prevent participants from knowing which condition they will be assigned to (e.g., "interested in how caring for something affects blood pressure"). However, participants must be fully informed via a consent form about what they will personally be doing, although not necessarily what other groups in the study are doing.
Timing of Measurement: For robust data, it is often beneficial to measure the dependent variable (e.g., blood pressure) both before (baseline) and after the experimental manipulation. More data points generally lead to better insights, though participant compliance must be considered. This provides a clearer picture of change.