Scribd
Upload
216 views

1.5 - Inference For A Single Proportion Using A Theory-Based Approach

1. The document discusses the theory-based approach for making inferences about a single proportion. It involves predicting the distribution under the null hypothesis, which is often bell-shaped. 2. At least 10 successes and 10 failures are needed for the distribution to be bell-shaped. The standard error and z-statistic formulas are provided. 3. An example comparing the theory-based approach to simulation is presented, showing the theory-based approach is invalid for small samples and does not always produce accurate results. Simulation should be used instead.

Uploaded by

Scarlett Ramos
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
216 views

1.5 - Inference For A Single Proportion Using A Theory-Based Approach

1. The document discusses the theory-based approach for making inferences about a single proportion. It involves predicting the distribution under the null hypothesis, which is often bell-shaped. 2. At least 10 successes and 10 failures are needed for the distribution to be bell-shaped. The standard error and z-statistic formulas are provided. 3. An example comparing the theory-based approach to simulation is presented, showing the theory-based approach is invalid for small samples and does not always produce accurate results. Simulation should be used instead.

Uploaded by

Scarlett Ramos
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 6

1.

5 Inference for a Single Proportion:


Theory-Based Approach

Theory-Based Approach
Used to predict what the distribution from the null
hypothesis will look like
o Will often a symmetric, bell-shaped curve
o Centered around the parameter value
Central Limit Theorem: for large sample sizes, the
distribution becomes approximately bell-shaped
Need at least 10 successes and 10 failures
(1)
Standard Error:

Standardized Statistic:

=
(1 )

Heart Transplant Example (from Section 1.3)
Using Simulation
o = .15
o = .197
o =361
o Simulated Standard Error = 0.019
. 197 .15
= = 2.47
0.019
Using Theory-Based Approach

0.197 0.15
= = = 2.5
(1 ) 0.15(1 0.15)
361
o Interpretation: Reject the null hypothesis because
z-score is bigger than 1.645
o Conclusion: The death rate at St. Georges is
higher than 15%
When Theory-Based Approach doesnt work
Example: Kickers make 76% of field goals from over 45
yards. One kicker makes 4 out of 8 field goals. Is he
worse than other kickers?
Number of successes = 4 (i.e. made field goal)
Number of failures =4 (i.e. missed field goal)
Theory-based approach doesnt work
because there isnt at least 10 successes
and 10 failures
Using Theory-Based approach (although its invalid):
H0 : = 0.76
HA : < 0.76
4
= = 0.5
8
0.5 0.76
= = = 1.72
(1 ) 0.76(1 .76)
8
-value = 0.0425

Interpretation: Reject the null hypothesis since p-


value = 0.0425 < 0.05
Conclusion: The long run proportion of made field
goals by this kicker is less than 76%.
Using Simulation

Interpretation: Fail to reject the null hypothesis


since p-value = 0.082 > 0.05
Conclusion: Theres not enough evidence that this
kickers long run proportion of made field goals is
less than 76%
Note: This contradicts the conclusion from the
theory-based approach. Why? Because the theory-
based approach is NOT valid!
Key Points
Need large sample size for theory-based approach
For small sample sizes, the distribution will not be
approximately bell-shaped => inaccurate standardized
statistic (z-score) and inaccurate p-value
Simulation ALWAYS works
Example
Is faking cell phone calls a common practice among cell
phone users? A report for International Business Times
wrote that 10% of cell phone users fake cell phone use.
To determine if this were true, the Pew Research Center
conducted a survey and asked a random sample of 1,858
American cell phone users and discovered that 13%
admitted to faking cell phone calls in the past 30 days.
1. Describe the parameter of interest in words.
Long run proportion of people that fake cell phone
calls

2. Write the hypotheses in symbols.


H0 : = 0.1
HA : 0.1

3. Explain if a theory-based approach is acceptable.


Yes, there are 13% 1858 = 242 people who faked
calls (more than 10) and there (1 13%) 1858 =
1616 people who did not fake calls (more than 10)

4. Calculate the standard error.


0.1(1 0.1)
= 0.007
1858
5. Calculate the standardized statistic.
0.13 0.1
= = = 4.29
0.007
(1 )

6. Give an interpretation and conclusion (use as


the cutoff level for the z-score).
Reject the null hypothesis because z > 2 and conclude
that the long run proportion of people that fake cell
phone calls is not 10%.

You might also like