Study Notes on Introduction to Statistics and Significance Testing
Introduction to Statistics: Significance Testing for a Population Proportion
Instructor: Josh Taber from Canyon Del Oro High School, Tucson, Arizona.
Topic: Introduction to statistics and the logic of significance testing for a population proportion.
Learning Objectives:
To understand how to identify evidence for a claim.
To determine if the evidence for a claim is convincing.
Example Case: Does Green Equal More Natural?
Background Context:
Companies often use green packaging to suggest healthiness or naturalness.
Research Focus: Do high school students associate the color green with being healthier or more natural?
Methodology of the Study:
Participants: 30 randomly selected students from a high school.
Experimental Design: Students taste two cups of lemonade (one green and one white) and indicate which tastes more natural.
Blindness: Both cups contained the same brand of lemonade, ensuring no bias from the flavor.
Results:
Out of 30 students, 18 (60%) preferred the lemonade from the green cup.
Analyzing the Evidence
Main Question: Does the preference for the green cup provide convincing evidence that students associate green with being more natural?
Expected Proportion without Effect:
If there was no effect, we would expect 50% to choose the green cup.
Observed Proportion (p̂):
The observed sample proportion, denoted as p̂, is given by:
p̂ = \frac{18}{30} = 0.60
Hypotheses: Two possible explanations for the results:
Random Chance Explanation: The preference for the green cup could have occurred by chance.
Real Effect Explanation: Students truly associate the color green with being more natural.
Testing the Evidence
Determining Convincing Evidence:
To assess the evidence for the second explanation, we need to calculate how likely it is to obtain a sample proportion of 60% or greater by chance, assuming the actual proportion is 50% (p = 0.5).
Simulation Method:
Use coin flips to simulate choosing between two cups (green and white).
Each flip represents a choice; record the number of heads (green cup choices).
Sample Simulation Results:
Example outcomes from simulated trials:
Trial 1: 14 out of 30 heads (p̂ = 0.467)
Trial 2: 17 out of 30 heads (p̂ = 0.567)
Trial 3: 12 out of 30 heads (p̂ = 0.400)
Continue until 100 trials are completed. Distribution typically centers around 50%.
Results of 100 Simulations:
Number of trials achieving 60% or more heads:
16 out of 100 trials (or 16%) showed p̂ ≥ 0.60.
Final Evaluation of Evidence
Revisiting Explanations:
Given that 16 out of 100 trials resulted in a sample proportion of 60% or higher by chance, we cannot rule out the random chance explanation.
Therefore, the claim that students associate the color green with being natural cannot be conclusively supported based solely on the experiment's data.
Conclusions
Key Takeaways:
Identifying Evidence for a Claim: Show results consistent with the claim.
Determining Convincing Evidence:
Consider both random chance and real effect as explanations for the observed results.
Estimate the probability of obtaining strong or stronger evidence by chance alone.
If random chance can be effectively ruled out, the evidence for the claim is considered convincing.
Future Learning: This logic will be revisited throughout units six, seven, eight, and nine in the program; mastery of this approach in reasoning and statistical thought is critical.