AP Statistics Unit 7: Inference for Quantitative Data - Means Study Guide

Topic 7.1: Introduction to Sampling Variability

• Concept of Sampling Variability: Samples vary naturally. A single sample mean ($\bar{x}$) is unlikely to exactly match the true population mean ($\mu$).
• Sampling Error: This natural variation between the sample mean and the population mean is termed "sampling variability" or "sampling error." It is not a mistake made by the researcher but a result of the random nature of sampling.
• Foundational Tools: Inference for means depends on constructing sampling distributions, which describe the behavior of all possible sample means.
• Normal vs. T Distribution:     * Sampling distributions for means are typically approximated using a normal distribution.     * In practice, the population standard deviation ($\sigma$) is rarely known. Instead, researchers use the sample standard deviation ($s$) as an estimate.     * When using $s$ to estimate $\sigma$, the T distribution is used because it provides a more accurate approximation.     * The T distribution is similar to the normal distribution but has "heavier" or larger tails (more area on the bottom and left/right ends).
• Degrees of Freedom ($df$):     * Every T distribution is defined by its degrees of freedom, calculated as:     $df = n - 1$     * As the sample size ($n$) increases, the degrees of freedom increase.     * As $df$ increases, the tails of the T distribution shrink, and the curve looks more and more like a standard normal distribution.

Topics 7.2 & 7.3: Constructing a Confidence Interval for a Population Mean

• Purpose: To estimate the true population mean ($\mu$) based on a sample mean ($\bar{x}$).

• The Four-Step Construction Process:     1. Step 1: Name and Context: Identify the procedure as a "One sample T interval for a population mean" and define the context of the parameter being estimated.     2. Step 2: Check Conditions:         * Randomness: The sample must be a random sample to avoid bias.         * 10% Rule: The sample size ($n$) must be less than 10% of the population to assume independence ($n < 0.10N$).         * Normal / Large Sample Condition: To ensure the sampling distribution of $\bar{x}$ is approximately normal, one of the following must be true:             * The population is known to be normal (any sample size works).             * The sample size is large ($n \ge 30$), satisfying the Central Limit Theorem (CLT).             * If $n < 30$ and the population distribution is unknown, the data must be checked (via histogram, dot plot, or box plot) to ensure there are no major outliers or significant skewness.     3. Step 3: Calculated Math:         * Formula: $\bar{x} \pm T^* \times \frac{s}{\sqrt{n}}$         * Critical Value ($T^<em>$): Found using a T table or the invT function on a calculator. Input the area to the left (e.g., for 95% confidence, use 0.025 at the bottom) and the $df$ ($n - 1$).          Standard Error: $\frac{s}{\sqrt{n}}$         * Margin of Error ($ME$): The product of the critical value and the standard error ($T^* \times \frac{s}{\sqrt{n}}$).     4. Step 4: Interpretation: Explicitly state: "I am [X]% confident that the true population mean of [Context] is between [Lower Bound] and [Upper Bound]."

• Factors Affecting Interval Width:     * Sample Size ($n$): Larger samples vary less. Increasing $n$ decreases the margin of error and narrows the interval.     * Confidence Level: Higher confidence levels (e.g., 99% vs. 95%) require a wider interval to capture the true mean in more possible samples.     * Ideal Balance: Researchers prefer high confidence combined with large sample sizes to achieve accurate, narrow intervals.

Example Case Study: AAA Batteries

• Scenario: Becky selects a random sample of $55$ batteries. The sample mean life is $348.5\,minutes$ with a standard deviation of $23.8\,minutes$.
• Goal: Construct a 99% confidence interval.
• Process:     * Step 1: One sample T interval for the mean life of name-brand AAA batteries in a specific flashlight.     * Step 2: Random sample of 55 batteries; $55 < 10\%$ of all batteries; $n = 55 \ge 30$ (CLT applies).     * Step 3: $df = 54$. Using invT (area $= 0.005$, $df = 54$), $T^* = 2.67$.     * Calculation: $348.5 \pm 2.67 \times \frac{23.8}{\sqrt{55}} = 348.5 \pm 8.5699$.     * Interval: $(340.13, 357.07)\,minutes$.     * Step 4: Becky is 99% confident the true population mean time the batteries last is between $340.13$ and $357.07\,minutes$.

Interpreting Confidence Level vs. Confidence Interval

• Common Error: Interpreting the confidence level as a probability (e.g., "there is a 95% chance the mean is here").
• Correct Definition: If we were to take many, many random samples and construct a confidence interval for each, approximately [X]% of those intervals would contain the true population mean ($\mu$).
• Logic: 95% of sample means are close enough to the center that their constructed intervals capture the true mean; the other 5% of intervals fail because they are based on extreme (unlikely) sample means.

Working Backwards: Determining Sample Size

• Sometimes researchers start with a desired margin of error ($ME$).
• Formula: $ME = Z^* \times \frac{s}{\sqrt{n}}$
• Note: Since we don't know the sample size, we can't find $df$ or $T^<em>$. In this case, we substitute the critical value $Z^</em>$ from a normal distribution to estimate.
• Example (Marcus and Marbles): Marcus wants a $ME$ of $0.4\,cm$ with 95% confidence and an estimated $s = 1.7\,cm$.     * $0.4 = 1.96 \times \frac{1.7}{\sqrt{n}}$     * Solving for $n$: $n \approx 70$. Marcus needs to sample at least $70$ marbles.

Topics 7.4 & 7.5: Significance Tests for a Population Mean

• Purpose: To test a claim about a population mean based on sample evidence.
• The Four-Step Significance Test Process:     1. Step 1: Hypotheses and Name:         * Name: One sample T test for a population mean.         * Null Hypothesis ($H_0$): The Status Quo ($\mu = \mu_0$).         * Alternative Hypothesis ($H_a$): The claim being investigated ($\mu > \mu_0$, $\mu < \mu_0$, or $\mu \neq \mu_0$).     2. Step 2: Conditions and Sampling Distribution:         * Verify Random, 10% Independence, and Normal/Large Sample conditions.         * State the Center ($\mu_{\bar{x}} = \mu_0$), Spread (Standard Error), and Shape (Approximate T distribution with $df = n - 1$).     3. Step 3: P-value and T-score Calculation:         * T-score: Measures how many standard errors the sample mean is from the null mean.         $t = \frac{\bar{x} - \mu_0}{s/\sqrt{n}}$         * P-value Definition: The probability of obtaining any sample mean as extreme or more extreme than the observed mean, assuming $H_0$ is true.         * Tool: tcdf on the calculator. Use $df$ and the lower/upper bounds defined by the T-score.         * Two-Sided Tests: If $H_a$ is $\mu \neq \mu_0$, double the P-value from the tail test.     4. Step 4: Conclusion:         * Compare P-value to significance level ($\alpha$).         * If P-value < $\alpha$: Reject $H_0$. There is significant evidence for $H_a$.         * If P-value > $\alpha$: Fail to reject $H_0$. There is not enough evidence for $H_a$.

Type I and Type II Errors and Power

• Type I Error: Rejecting the null hypothesis ($H_0$) when it is actually true (a false positive).
• Type II Error: Failing to reject the null hypothesis ($H_0$) when the alternative hypothesis ($H_a$) is actually true (a false negative).
• Power: The probability of correctly rejecting a false null hypothesis.     * Power is desirable.     * To Increase Power: Increase the sample size ($n$). Larger samples are more accurate to the truth and better at detecting differences.

Topics 7.6 & 7.7: Confidence Intervals for the Difference of Two Means

• Purpose: To estimate the difference between two population means ($\mu_1 - \mu_2$).
• Formulas:     * Point Estimate: Difference in sample means ($\bar{x}_1 - \bar{x}_2$).     * Standard Error: $\sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}$     * Interval: $(\bar{x}_1 - \bar{x}_2) \pm T^* \times \sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}$
• Degrees of Freedom for Two Samples:     1. Use the complex calculator formula (most accurate).     2. Use the smaller of the two individual degrees of freedom values ($df = \min(n_1-1, n_2-1)$).     3. Sum the degrees of freedom ($df_1 + df_2$) (less accurate but often accepted on the AP exam).
• Interpretation:     * If the interval is entirely positive (e.g., 1.2 to 14.0), group 1 is significantly larger than group 2.     * If the interval is entirely negative, group 1 is significantly smaller than group 2.     * If the interval contains Zero, there is no significant evidence of a difference between the means.

Topics 7.8 & 7.9: Significance Tests for the Difference of Two Means

• Name: Two sample T test for the difference between population means.
• Hypotheses:     * $H_0: \mu_1 = \mu_2$, or $\mu_1 - \mu_2 = 0$.     * $H_a: \mu_1 > \mu_2$, $\mu_1 < \mu_2$, or $\mu_1 \neq \mu_2$.
• T-score Calculation for Differences:$t = \frac{(\bar{x}_1 - \bar{x}_2) - 0}{\sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}}$
• Example Case Study: Resting Heart Rate:     * Scenario: 10 exercisers ($\bar{x} = 60.1$) vs. 8 non-exercisers ($\bar{x} = 79.625$).     * H1: $H_0: \mu_{ex} = \mu_{no}$; $H_a: \mu_{ex} < \mu_{no}$.     * Standard Error: $2.8369$.     * T-score: $-6.88$.     * P-value: $0.000175$.     * Conclusion: Since the P-value ($0.000175$) is less than $\alpha = 0.01$, we reject $H_0$. There is significant evidence hikers/exercisers have a lower mean resting heart rate.