Confidence Intervals Calculations

Why This Matters

Confidence intervals are the bridge between your sample data and the truth about an entire population—and that's exactly what you're being tested on. Every time you calculate a CI, you're quantifying uncertainty, which is the heart of statistical inference. The AP exam loves testing whether you understand why intervals widen or narrow, when to use z versus t distributions, and how to interpret what a 95% confidence level actually means (hint: it's not "95% chance the parameter is in this interval").

The concepts here connect directly to hypothesis testing, sampling distributions, and the Central Limit Theorem. You'll need to recognize when conditions are met, choose the right formula, and—critically—interpret results in context. Don't just memorize formulas; know what each component controls and how changing sample size, confidence level, or variability affects your interval. Master the mechanics behind these calculations, and FRQs become much more manageable.

---

Foundational Concepts: What CIs Actually Measure

Before diving into formulas, you need to understand what confidence intervals represent. A CI captures the range of plausible values for a population parameter based on sample data, with a specified level of confidence.

Definition and Purpose

• A confidence interval estimates a range where the true population parameter likely falls—it's not a probability statement about any single interval
• The width reflects uncertainty—wider intervals mean less precision, narrower intervals mean more confidence in your estimate
• CIs quantify sampling variability, connecting your sample statistic to the unknown population parameter through probability

Confidence Level and Margin of Error

• The confidence level (e.g., 95%) means that if you repeated the sampling process many times, about 95% of the resulting intervals would contain the true parameter
• Margin of error is the "±" portion—calculated as $\text{critical value} \times \text{standard error}$
• Higher confidence = wider interval, because you need a larger range to be "more sure" you've captured the parameter

---

Single-Parameter Estimation: Means and Proportions

These are your bread-and-butter calculations. The choice of formula depends on what you're estimating and what information you have about the population.

CI for Population Mean (Known $\sigma$)

• Formula: $\bar{x} \pm z^* \cdot \frac{\sigma}{\sqrt{n}}$ where $\bar{x}$ is the sample mean, $\sigma$ is the known population standard deviation, and $n$ is sample size
• Use z-scores for critical values (e.g., $z^* = 1.96$ for 95% confidence)—this assumes normality or $n \geq 30$ via the Central Limit Theorem
• Rarely used in practice because knowing $\sigma$ without knowing $\mu$ is uncommon—but it's a foundational concept for exams

CI for Population Mean (Unknown $\sigma$)

• Formula: $\bar{x} \pm t^* \cdot \frac{s}{\sqrt{n}}$ where $s$ is the sample standard deviation and $t^*$ comes from the t-distribution with $df = n - 1$
• The t-distribution has heavier tails than the normal distribution, accounting for the extra uncertainty when estimating $\sigma$ from sample data
• As $n$ increases, the t-distribution approaches the normal distribution—this is why large samples can use either method

CI for Population Proportion

• Formula: $\hat{p} \pm z^* \cdot \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$ where $\hat{p}$ is the sample proportion
• Conditions required: $n\hat{p} \geq 10$ and $n(1-\hat{p}) \geq 10$ for the normal approximation to be valid
• The standard error $\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$ is maximized when $\hat{p} = 0.5$, which is why conservative sample size calculations use this value

---

Factors That Control Interval Width

Understanding these relationships is crucial for exam questions that ask "what happens if...?" Interval width is determined by three factors: confidence level, sample size, and variability.

Effect of Sample Size

• Increasing $n$ decreases interval width because $\sqrt{n}$ appears in the denominator of the standard error
• The relationship is square root, so quadrupling your sample size only halves the margin of error
• Larger samples provide more information, reducing the uncertainty in your estimate of the population parameter

Relationship Between Confidence Level and Width

• Higher confidence levels require larger critical values ($z^*$ or $t^*$), which directly increases the margin of error
• 99% CI is wider than 95% CI for the same data—you're trading precision for confidence
• Choosing a confidence level involves balancing the need for certainty against the desire for a useful (narrow) interval

Sample Size Determination

• For means: $n = \left(\frac{z^* \cdot \sigma}{E}\right)^2$ where $E$ is the desired margin of error
• For proportions: $n = \frac{(z^*)^2 \cdot \hat{p}(1-\hat{p})}{E^2}$—use $\hat{p} = 0.5$ if no prior estimate exists
• Always round up when calculating required sample size—you can't survey 384.2 people

---

Comparing Two Populations

When you need to compare groups rather than estimate a single parameter, the formulas combine information from both samples. The standard error for a difference involves adding variances, not standard deviations.

Difference Between Two Means

• Formula: $(\bar{x}_1 - \bar{x}_2) \pm t^* \cdot \sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}$ for independent samples
• Degrees of freedom can be calculated using Welch's approximation or conservatively as $\min(n_1-1, n_2-1)$
• If the interval contains zero, you cannot conclude a significant difference between the population means at that confidence level

Difference Between Two Proportions

• Formula: $(\hat{p}_1 - \hat{p}_2) \pm z^* \cdot \sqrt{\frac{\hat{p}_1(1-\hat{p}_1)}{n_1} + \frac{\hat{p}_2(1-\hat{p}_2)}{n_2}}$
• Conditions: Both samples must independently satisfy $n\hat{p} \geq 10$ and $n(1-\hat{p}) \geq 10$
• Interpretation: If zero is not in the interval, evidence suggests a real difference between population proportions

---

Special Cases and Advanced Methods

These topics extend the basic framework to handle specific situations. Knowing when to apply each method demonstrates deeper statistical understanding.

One-Sided vs. Two-Sided Intervals

• Two-sided CIs estimate parameters in both directions (e.g., $\mu$ could be higher or lower than $\bar{x}$)
• One-sided CIs bound the parameter in only one direction—used when you only care about exceeding or falling below a threshold
• Critical values differ: a one-sided 95% CI uses $z^* = 1.645$, while two-sided uses $z^* = 1.96$

Use of t-Distribution for Small Samples

• Required when $n < 30$ and population standard deviation is unknown—the CLT doesn't guarantee normality
• Heavier tails compensate for the increased variability in estimating $\sigma$ from small samples
• Check the population distribution: if the underlying data is heavily skewed, even t-intervals may be unreliable for small $n$

Prediction Intervals

• Formula: $\bar{x} \pm t^* \cdot s\sqrt{1 + \frac{1}{n}}$ estimates where a single new observation will fall
• Wider than confidence intervals because they account for both sampling variability and individual observation variability
• The "1 +" term captures the additional uncertainty of predicting one value rather than estimating a mean

---

Assumptions and Validity Conditions

No confidence interval is valid unless conditions are met. Checking assumptions isn't optional—it's required for full credit on FRQs.

Required Conditions

• Random sampling: Data must be randomly selected from the population of interest
• Independence: Observations must be independent; for sampling without replacement, check that $n \leq 0.10N$ (10% condition)
• Normality/sample size: For means, either the population is normal or $n \geq 30$; for proportions, check $n\hat{p} \geq 10$ and $n(1-\hat{p}) \geq 10$

Bootstrap Method

• Resampling with replacement creates thousands of simulated samples from your original data
• No normality assumption required—the bootstrap builds an empirical sampling distribution
• Particularly useful for complex statistics, small samples, or when traditional assumptions fail

---

Quick Reference Table

---

Self-Check Questions

1. When constructing a CI for a population mean, what two conditions determine whether you use a z-interval or a t-interval?

2. Compare and contrast: How does increasing the confidence level from 90% to 99% affect interval width, and how does quadrupling the sample size affect it? Which change narrows the interval?

3. A confidence interval for $\mu_1 - \mu_2$ is $(-2.3, 4.7)$. What conclusion can you draw about whether the population means differ? Explain your reasoning.

4. Which two formulas both contain the term $\sqrt{\hat{p}(1-\hat{p})}$, and why does this expression appear in both?

5. An FRQ asks you to construct a 95% CI for a proportion with $n = 40$ and $\hat{p} = 0.08$. Before calculating, what condition should you check, and is it satisfied? What are the implications?