Unit 3 Lesson 4 Experimental Design and Confounding
Experimental Design and Confounding
Unit 3, Topics 3.6–3.7: Experimental Design and Confounding
Overview
This lesson focuses on designing experiments to control for confounding variables, which are outside factors that affect both the explanatory variable (the one manipulated, like a treatment) and the response variable (the outcome measured, like growth rate), making it hard to isolate the true effect. Blocking groups participants by a known confounder (e.g., soil type in a fertilizer test) to compare within similar groups, reducing its influence and improving accuracy.
Example: In a teaching app trial, prior knowledge confounds by affecting app engagement (explanatory) and learning speed (response); blocking by skill level (beginner/advanced) compares within groups to reveal the app's true boost.
Context, such as the population and real-world constraints, is vital because it guides design choices (e.g., blocking by age in a drug trial ensures fair comparisons across life stages, but ignoring ethics like consent could undermine results).
Example: For a fitness tracker study on office workers, blocking by desk job vs. active role fits the population's constraints, but skipping consent risks invalid data from unwilling participants.
By identifying confounders and using blocking, experiments become more valid for causation claims, though risks like unknown confounders still require careful interpretation.
Example: Blocking a coffee vs. tea alertness test by caffeine tolerance strengthens claims of "Coffee causes sharper focus," but unblocked factors like sugar intake might still obscure full effects.
Assignment:
Part 1: Guided Practice Activity
Consider the example of testing a new fertilizer's effect on crop yield, with 40 plots varying in soil type (sandy vs. clay). Follow the tasks below to design a blocked experiment and address confounding.
Example Scenario: Apply fertilizer to half the plots and none to the other half, then measure yield in bushels per acre.
Tasks:
- Identifying Confounding:
- Define confounding as a factor that influences both the explanatory variable (fertilizer) and response variable (yield), obscuring the true effect.
Example: Soil type is a confounder because it affects how well fertilizer works (explanatory) and natural yield (response), so sandy plots might yield more regardless of fertilizer, hiding its real impact. - Write 1-2 sentences explaining how it obscures results in the scenario.
Example: Confounding from soil type obscures the fertilizer's effect by making clay plots seem less responsive even if the fertilizer works equally, leading to false claims about its uselessness on certain soils. - Extra Practice: For a painkiller trial on headaches, identify a confounder like stress level and explain its obscuring effect.
- Define confounding as a factor that influences both the explanatory variable (fertilizer) and response variable (yield), obscuring the true effect.
- Designing Experiments with Blocking:
- Describe blocking as grouping by a confounder (e.g., soil type) and comparing treatments within each block to account for its influence.
Example: Divide 40 plots into two blocks of 20 sandy and 20 clay, then randomly assign fertilizer to 10 in each block—this isolates the fertilizer's effect by comparing within similar soils. - Design a blocked experiment for the fertilizer scenario, including steps for randomization and replication.
Example: Block by soil (20 sandy, 20 clay); within each, randomly assign 10 to fertilizer/10 to control using coin flips; replicate by measuring yield over two harvests to check consistency. - Extra Practice: For your painkiller trial, design a blocked experiment by headache severity.
- Describe blocking as grouping by a confounder (e.g., soil type) and comparing treatments within each block to account for its influence.
- Justifying Choices and Reflection:
- Justify your design choices with 1-2 sentences, linking to reducing confounding (e.g., "Blocking by soil ensures comparisons are fair within types, minimizing the confounder's obscuring role.").
Example: Random assignment within blocks justifies fairness, as it balances other factors like sunlight, improving validity for claiming fertilizer boosts yield by 20% regardless of soil. - Write 2-3 sentences interpreting sample results (e.g., 15% higher yield in fertilizer plots) while considering confounding risks.
Example: The 15% yield increase suggests fertilizer works, but if unblocked soil confounds remain (e.g., uneven watering), results might overstate benefits; in farmer context, this risk means testing on mixed fields before full use.
- Justify your design choices with 1-2 sentences, linking to reducing confounding (e.g., "Blocking by soil ensures comparisons are fair within types, minimizing the confounder's obscuring role.").
Part 2: Independent Practice
Design an experiment to test if a new app improves vocabulary learning, using 30 students with varying English levels (beginner vs. advanced). Measure words learned per week.
Tasks:
- Identify a potential confounder (e.g., prior English exposure) and explain how it obscures the app's effect on learning.
- Design a blocked experiment, including steps for grouping, assignment, and replication.
- Justify your design choices in 1-2 sentences, focusing on reducing confounding.
- Write 2-3 sentences interpreting hypothetical results (e.g., 10% more words learned with app) while noting remaining confounding risks.
- Extra Activity: Invent your own experiment (e.g., testing light exposure on plant height). Identify a confounder, design with blocking, justify, and interpret results.
Homework Assignment
- Brainstorm a real-world experiment (e.g., comparing two study techniques for memory retention). Identify a confounder, design a blocked experiment with steps, justify choices, and interpret potential results with risks to share next class.