Introduction to Hypothesis Testing and the One-Sample Z-Test

Importance of Chapter 8: This chapter represents the absolute "heart" of the course.
Building Blocks: All material covered in chapters one through seven serves as the necessary building blocks to reach this point.
Future Course Outlook: Every single topic and chapter following Chapter 8 will revolve around hypothesis testing. Each subsequent chapter will present different types of hypothesis tests tailored to specific situations, research designs, and study types.
Introductory Focus: Today’s lecture focuses on hypothesis testing in general through a specific example called the one-sample z-test.

Course Packet Handouts: The lecture follows the "Chapter 8 Day 1 Lecture Handout" found on page 75 of the course packet (based on the Fall 2020 version).
Essential Handouts:
- Steps for One-Sample Z-Test: Primarily useful for laboratory assignments.
- Hypothesis Test Reminders Handout: Described as the most important handout in the class. It provides general principles that apply to all hypothesis testing topics for the remainder of the semester.
Study Recommendation: Because hypothesis testing is the foundation for everything moving forward, students are advised to read the textbook for Chapter 8 at least three times. This aligns with the three scheduled instruction days (Day 1, Day 2, and the Extra Day).

Hypothesis Testing Foundation: Hypothesis testing is built directly upon the concept of the distribution of sample means developed in Chapter 7.
Definition: The distribution of sample means is created by taking all possible samples of size $n$ (where $n$ is the number of scores in each sample) from a population and calculating the mean ( $M$ ) for each.
Scale of Samples: In real-world populations (e.g., millions of students), the number of possible samples is unmanageable (millions, billions, or trillions). This makes the Central Limit Theorem essential for practical application.
Key Symbols:
- Sample Mean: $M$
- Population Mean: $\mu$
- Standard Deviation: $\sigma$

The CLT identifies three critical characteristics of the distribution of sample means:

The Center (Central Tendency):
- Known as the Expected Value of M.
- Regardless of the parent population size, the mean of all sample means ( $M$ ) will always equal the mean of the parent population ( $\mu$ ).
- Formula: $E(M) = \mu$
The Width (Variability):
- Known as Standard Error ( $\sigma_M$ ). It is essentially the standard deviation of a distribution of means.
- Relates the spread of sample means back to the parent population standard deviation divided by the square root of the sample size.
- Formula: $\sigma_M = \frac{\sigma}{\sqrt{n}}$
The Shape:
- The distribution is considered "Normal" if at least one of two conditions is met:
  - The parent population is normal.
  - The sample size ( $n$ ) is at least 25 to 30 (a common rule of thumb).
- Statistical accuracy increases as $n$ approaches infinity, though 25-30 is sufficient for most research purposes.

Scenario: Chaffey College Research Office data shows the average age of all students ( $\mu$ ) is $26\,\text{years}$ with a standard deviation ( $\sigma$ ) of $6\,\text{years}$ .
Research Question: What is the probability that a random sample of $36$ students ( $n = 36$ ) will have a mean age ( $M$ ) greater than $30$ ?
Calculation Steps:
- Identify Distribution: Since $n=36$ , the distribution of sample means is normal.
- Calculate Standard Error: $\sigma_M = \frac{6}{\sqrt{36}} = \frac{6}{6} = 1$
- Calculate Z-score: $z = \frac{30 - 26}{1} = +4.00$
- Determine Probability: Looking up a z-score of $4.00$ in the unit normal table (column C for the tail) gives a proportion of $0.00003$ .
- Percent chance: $0.003\%$
Interpretation: This result is an extreme outlier. In a world where you drew 100,000 such samples, only 3 would have a mean age above 30. If this actually happened, you would be extremely surprised; it is equivalent to winning the lottery.

The Tweak: Imagine we do not know the actual $\mu$ and $\sigma$ for Chaffey College. Instead, we look at Mt. San Antonio College (Mt. SAC), which reports $\mu = 26$ and $\sigma = 6$ .
The Starting Assumption: We assume Chaffey students are identical to Mt. SAC students ( $\mu = 26$ , $\sigma = 6$ ).
The Result: If we draw a sample of 36 Chaffey students and get a mean of 30, the math remains the same ( $z = 4.00$ , $p = 0.003\%$ ). However, the logic shifts from absolute to conditional.
Conditionality: The probability is only $0.003\%$ if our assumption about Chaffey being like Mt. SAC is true.

When a researcher observes an extremely unlikely result (low probability), they are faced with two logical choices:

Stick to the Assumption: Maintain that the starting assumption (e.g., Chaffey = Mt. SAC) is true and conclude that an incredibly rare, "miraculous" event simply occurred.
Reject the Assumption: Conclude that the starting assumption was likely incorrect. If the actual mean for Chaffey was higher (e.g., 28 or 29), finding a sample mean of 30 would not have been nearly as unlikely.

The Weatherman: If a meteorologist says there is a $1\%$ chance of rain and it pours, you can either believe the weatherman was right and a rare event happened, or conclude the weatherman's model (the starting assumption) was wrong.
The 2016 Election: Polls gave one candidate a $90\%$ chance and the other a $10\%$ chance. When the underdog won, critics either claimed the polls were fundamentally flawed (rejected the assumption) or recognized that a $10\%$ event happened (unlikely things do occur).
Sports Betting: When a heavy underdog wins, it represents either a failure in the odds-making logic or the occurrence of a low-probability miraculous victory.