TPS5e_Ch4.2

Chapter 4: Designing Studies (Section 4.2: Experiments)

Page 1: Introduction

• Focus on the topic of experiments within the context of statistics.

• Key source: "The Practice of Statistics, 5th Edition" by Starnes, Tabor, Yates, Moore.

Page 2: Learning Objectives

• Distinguish between an observational study and an experiment.

• Explain the concept of confounding.

• Identify the following in an experiment:

  • Experimental units

  • Explanatory variables

  • Response variables

  • Treatments

• Explain the roles of:

  • Comparison

  • Random assignment

  • Control

  • Replication

• Describe a completely randomized design.

• Describe the placebo effect and purpose of blinding in experiments.

• Interpret statistical significance in the context of experiments.

• Explain blocking in experiments.

• Describe randomized block design and matched pairs design.

Page 3: Observational Study vs. Experiment

• Observational Study:

  • Observes individuals and measures variables without influencing responses.

• Experiment:

  • Deliberately imposes treatments on individuals to measure responses.

• Significance: Experiments are essential for understanding cause and effect.

• Key distinction between studies is critical in statistics.

Page 4: Confounding

• Confounding: Occurs when two variables are associated such that their individual effects on a response cannot be differentiated.

• Observational studies often suffer from confounding, leading to misinterpretation.

• Importance: Understanding the root cause of responses is crucial in experiments.

Page 5: The Language of Experiments

• Experiment: Involves imposing treatments to observe responses.

• Treatment: A specific condition applied to individuals in an experiment.

• Experimental Units: The smallest collection of individuals that treatments are applied to. Often referred to as "subjects" when human beings are involved.

Page 6: Designing Failed Experiments

• Laboratory experiments may succeed with simple designs.

• Field experiments and studies involving animals or humans face more challenges due to variability.

• Bad designs lead to confounding, resulting in ineffective or misleading outcomes.

Page 7: Conducting Proper Experiments

• Solution to confounding: Implement a comparative experiment with different treatment groups.

• Random Assignment: Assigning experimental units to treatments randomly to minimize bias.

Page 8: Principles of Experimental Design

1. Comparison: Compare two or more treatments.

2. Random Assignment: Assign units to treatments by chance to create equivalent groups.

3. Control: Keep other variables constant across all groups.

4. Replication: Use enough experimental units for reliable results and differentiation away from chance variations.

Page 9: Completely Randomized Design

• In a completely randomized design, treatment assignments are based on chance.

• May involve a control group receiving either an inactive or standard treatment.

Page 10: Placebo Effect & Blinding

• Placebo Effect: Response to an inactive treatment.

• Double-Blind Experiment: Neither subjects nor research staff know which treatment the subjects receive.

• Maintaining consistency for all subjects is vital for valid results.

Page 11: Statistically Significant Results

• Statistically Significant: An outcome that is unlikely to occur by chance, implying causation.

• Experimental studies aim to observe significant response differences beyond random chance.

Page 12: Blocking

• Blocking: Groups of similar individuals are organized to improve estimates in experiments.

• In a randomized block design, assignments are made separately for each block to ensure better treatment comparisons.

Page 13: Matched Pairs Design

• Common method to compare two treatments involving pairs of similar experimental units, known as matched pairs design.

• Each unit in the pair can receive both treatments with the sequence randomized to avoid order effects.

Page 14: Section Summary

• Key points reviewed:

  • Distinguishing observational studies from experiments.

  • Understanding confounding variables and their implications.

  • Identification of crucial elements in experimental design including controls, treatments, randomization, and blinding.

  • Interpretation of statistical significance in relation to experimental results.