Teacher Chapter 12 Lesson 1 (1)

Chapter 12 Lesson 1: Experiments and Observational Studies

Observational Studies

• Observational Study: A method of gathering data without manipulating or requesting changes in the subjects.

  • Retrospective Study: Involves obtaining data from historical records.

  • Prospective Study: Involves gathering data in real-time as events happen.

• Key Point: Observational studies cannot establish causation—it only reveals correlations.

Experiments

• Experiment: Imposes treatments on individuals and observes outcomes.

  • Requires random assignment of experimental units to treatment groups for validity.

  • Factor: An explanatory variable that may affect outcomes.

    • Levels: Specific values assigned to the factor.

  • Blocking: A methodology in experiments to control variability.

• Key Point: Controlled experiments can establish causation due to the manipulation of variables.

Difference Between Observational Study and Experiments

• Example Case: Involves a bank surveying its 60 employees.

  • Options:

    • (A) Should not be used because it's observational.

    • (B) Cannot prove causation.

    • (C) Lacks random sampling, weakening results.

    • (D) Confidence intervals needed for estimates.

    • (E) No inference procedure required since the survey is a census.

Designing Experiments: Driving Simulation Example

• Research Question: Does using hands-free devices distract drivers?

• Important Variables:

  • Influencing factors: age, experience, emotional state, car type, vision, etc.

• Solution: Use random assignment to create two equal groups rather than blocking all variables due to complexity.

Four Principles of Experimental Design

1. Control: Manage variables to maintain consistency (e.g., all subjects in similar conditions).

2. Randomize: Randomly assign individuals to groups to minimize variability.

3. Replicate: Perform experiments with multiple subjects to verify results.

4. Block: Use blocking to reduce variability if applicable.

Practical Application: Driving Simulation Study

• Conducted with 78 university students in a simulator to analyze effects of cognitive load (hands-free conversation) on braking behavior.

  • Groups: Control (no conversation) vs. Experimental (with phone call).

  • Findings: Distracted drivers had different speeds and stopping distances compared to non-distracted drivers.

Understanding Experiment Validity

• New Headache Remedy Experiment

  • Out of 25 subjects, 20 reported relief after four hours. Can we conclude effectiveness?

    • Options: a) Effective; b) Inconclusive due to small sample; c) No control group; d) Superior to aspirin; e) Ineffective.

Case Study on Cover Design Sales

• CD Manufacturer Experiment studied sales performance with two different cover designs.

  • Random assignment with significant sales differences noted between designs.

    • Conclusion: Difference may indicate design impact, with considerations of random assignment in stores.

Pfizer COVID-19 Vaccine Trial Summary

• Ongoing trial with a random assignment of subjects aged 16 and over to test vaccine efficacy vs. placebo over 21 days.

  • Key Details: Vaccine formulation and efficacy endpoints regarding COVID-19.

Steps to Explain an Experiment

1. Describe Subjects/Units: Infer results only to this population.

2. Describe Treatments: Include all treatment factors and levels.

3. Random Assignment: Assign subjects to treatment groups randomly.

4. Descriptive Analysis: State outcomes to be analyzed (with units).

Group Work Example: Shrimp Growth Study

• Biologist will test the effect of nutrients and salinity on shrimp growth using 12 tanks with random placement.

  • Treatments include varying nutrients and salinity levels.

  • Statistical advantages and disadvantages of using only one shrimp type.

Blinding in Experiments

• Purpose: Avoid bias from knowledge of treatment among subjects and evaluators.

  • Types:

    • Single Blind: Only one group is unaware.

    • Double Blind: Both groups (treatment and evaluators) are unaware.

Placebo in Experiments

• Definition: A fake treatment mimicking actual treatments used to isolate the effect of the treatment.

• Placebo Effect: Changes resulting from the subject's belief in the treatment rather than the treatment itself.

Statistical Significance

• A result is statistically significant if it's unlikely to have occurred by chance alone, suggesting a real effect.

Variables in Experiments

• Lurking Variables: Factors that might influence results without being directly measured (example: relationship between firefighters and damage costs).

• Confounding Variables: Uncontrolled variables affecting response outcomes (example: cholesterol reduction could stem from a drug or lifestyle change).

Group Work: Alzheimer's Study Analysis

• Analyze the relationship between smoking and Alzheimer's risk based on tracked medical histories over 23 years.

  • Identify response and explanatory variables, and evaluate whether it was an observational study or an experiment, discussing possible confounding factors like exercise.