2 proportion z test
Introduction to Proportional Hypothesis Testing
When comparing two proportions, choose p1 (the proportion of one group) to be larger than p2 (the proportion of another group).
This simplifies calculations, as p1 - p2 will be a positive value.
Set the null hypothesis as H0: p1 - p2 = 0, indicating no difference between the two proportions.
Null Hypothesis and Assumptions
Null hypothesis (H0) states that p1 = p2, implying both proportions derive from the same population.
Verify the assumption that n1 and n2 are large enough (greater than 10).
Combine the sample sizes for the pooled proportion: p-hat pooled = (x1 + x2) / (n1 + n2).
Pooled Proportion Calculation
Pooled proportion should account for individual sample sizes:
p-hat pooled = (x1 + x2) / (n1 + n2).
Use this to ensure adequate sample representation when comparing proportions.
Check conditions:
n1 * p-hat pooled >= 10
n1 * q-hat pooled >= 10
n2 * p-hat pooled >= 10
n2 * q-hat pooled >= 10
Statistical Test
For hypothesis testing, conduct a two-sample Z test for proportions:
Test statistic formula: Z = (p-hat1 - p-hat2 - 0) / std_dev(p-hat1 - p-hat2).
Determine the Z score which indicates how many standard deviations away the sample proportion difference lies from the null hypothesis value.
Active vs Passive Solar Heating Systems
Differentiate between two groups in the solar heating context:
Passive Solar Heating Systems: The house itself is designed to collect solar energy (e.g., adobe houses).
Active Solar Heating Systems: Use mechanical devices, like solar panels, to convert sunlight into energy efficiently.
Evaluate the proportion of homes in each group that require less than 200 gallons of oil for fuel in a year.
Identifying Hypothesis and Data
Determine if it’s a proportion problem (not a means problem) and decide on a two-sample hypothesis test:
H0: p1 (active) - p2 (passive) not equal to.
Sample sizes from passive (n1) and active (n2) groups must be confirmed to be random and meet the size conditions.
Test Procedure
Example data:
For passive systems: 46 out of 52 homes require less than 200 gallons of oil.
For active systems: 37 out of 54 homes require less than 200 gallons of oil.
Compute pooled proportion:
p-hat pooled = (46 + 37) / (52 + 54) = 78.3%.
Validate n1 * p-hat pooled and n2 * p-hat pooled for both samples.
Conclusion and Decision Making
Calculate the Z score based on proportions and use it to derive the p-value.
Compare p-value with the chosen alpha (e.g., 1%, 5%):
If p-value > alpha, fail to reject the null hypothesis (H0).
Interpret results: If H0 is not rejected, the proportions are statistically similar, suggesting no significant difference in fuel consumption.
Summary of Findings
Reflect on the experimental outcomes:
If the observed proportion difference (~19.9%) is likely due to sampling variation (1.28% of the time), the active and passive systems do not differ significantly in oil consumption.
Acknowledge the larger context of how proportions in energy-efficient homes can inform building practices and energy conservation efforts.