Welcome to Unit 3: Inference for Proportions!
In Unit 1 and Unit 2, we spent a lot of time looking at data we already had in our hands. Now, we are entering the world of Statistical Inference. This is where the magic happens: we use a small group (a sample) to make smart guesses about a much larger group (a population).
Think of this chapter as the "bridge" between probability and making real-world claims. We are going to learn how sample proportions behave and why they are so reliable—if we follow the rules!
3.1 Estimators: Statistics vs. Parameters
Before we can dive into the math, we need to be very careful with our vocabulary. In statistics, we use statistics to estimate parameters.
Parameters vs. Statistics
- Parameter (\(p\)): The true value for the entire population. This is usually a "mystery" we are trying to solve.
- Statistic (\(\hat{p}\)): The value calculated from our sample. We use this to estimate the parameter. We call this a point estimator.
Bias and Variability
Imagine you are an archer shooting at a target. The bullseye is the true population parameter (\(p\)).
1. Bias (Accuracy): An estimator is unbiased if the center of its sampling distribution is equal to the true parameter. If your "aim" is consistently off to the left, your estimator is biased.
2. Variability (Precision): This describes how spread out your shots are. Large samples produce lower variability than small samples. If your shots are all over the place, you have high variability.
The Goal: We want an estimator with low bias (accurate) and low variability (precise)!
Quick Review:
- Parameter: The "Truth" (Population)
- Statistic: The "Estimate" (Sample)
3.2 The Sampling Distribution for \(\hat{p}\)
What if you took a sample of 100 students and found that \(25\%\) liked broccoli? Then, you took another sample of 100 students and found \(28\%\) liked broccoli? If you did this thousands of times and graphed all those \(\hat{p}\) values, you would have a Sampling Distribution.
The sampling distribution of \(\hat{p}\) tells us what values the sample proportion takes in repeated samples and how often it takes those values.
The "Three Pillars" of the Sampling Distribution
To describe a sampling distribution, we need to know its Center, Spread, and Shape.
1. The Center (The Mean)
The mean of the sampling distribution of \(\hat{p}\) is exactly equal to the population parameter \(p\). This is why we say \(\hat{p}\) is an unbiased estimator of \(p\).
\(\mu_{\hat{p}} = p\)
2. The Spread (The Standard Deviation)
The standard deviation measures how much the sample proportions typically vary from the true \(p\). Notice that as the sample size (\(n\)) gets bigger, the spread gets smaller!
\(\sigma_{\hat{p}} = \sqrt{\frac{p(1-p)}{n}}\)
Note: This formula only works if the 10% Condition is met (see below).
3. The Shape (The Normal Approximation)
The sampling distribution of \(\hat{p}\) will look like a beautiful, symmetric Normal curve if the sample size is large enough. We check this using the Large Counts Condition.
Did you know? The size of the population doesn't actually matter for the spread! A sample of 1,000 people from a city of 50,000 gives you just as much precision as a sample of 1,000 people from a country of 300 million.
The Three Mandatory Conditions
Whenever you are working with sampling distributions for proportions, you must verify these three conditions. If these aren't met, your results might be "statistically garbage."
- Randomization Condition: The data must come from a random sample or a randomized experiment. This helps us avoid bias.
- 10% Condition: When sampling without replacement, your sample size (\(n\)) must be less than \(10\%\) of the total population (\(N\)). This allows us to use the standard deviation formula as if the selections were independent.
\(n \leq 0.10N\) - Large Counts Condition (Normality): To use the Normal distribution for our shape, we need to expect at least 10 "successes" and at least 10 "failures" in our sample.
\(np \geq 10\) and \(n(1 - p) \geq 10\)
Memory Tip: Think of the 10% condition as the "Small Enough" rule (so we don't change the population too much) and the Large Counts condition as the "Big Enough" rule (so we get a Normal shape)!
Step-by-Step: Solving Problems
If an AP question asks you to "Describe the sampling distribution" or find the probability of a certain \(\hat{p}\), follow these steps:
Step 1: Identify the parameters. Find \(p\) (population proportion) and \(n\) (sample size).
Step 2: Check conditions. Mention Random, 10%, and Large Counts explicitly. Example: "Because \(np = 100(0.2) = 20 \geq 10\), the distribution is approximately Normal."
Step 3: Calculate the mean and standard deviation. Use \(\mu_{\hat{p}} = p\) and \(\sigma_{\hat{p}} = \sqrt{\frac{p(1-p)}{n}}\).
Step 4: Calculate the z-score (if needed). To find a probability, turn your \(\hat{p}\) into a z-score:
\(z = \frac{\text{statistic} - \text{parameter}}{\text{standard deviation}} = \frac{\hat{p} - p}{\sigma_{\hat{p}}}\)
Step 5: Find the probability. Use your calculator or Table A to find the area under the Normal curve.
Common Mistakes to Avoid
- Mixing up \(p\) and \(\hat{p}\): Remember, \(p\) is the "Truth" from the problem's setup. \(\hat{p}\) is what happened in one specific sample.
- Forgetting the 10% Rule: If you don't check this, you technically aren't allowed to use the standard deviation formula.
- Using \(n\) in the Large Counts check: Make sure you multiply \(n \times p\). Just having a large \(n\) isn't enough if \(p\) is extremely tiny (like \(0.0001\)).
- Language errors: Never say the distribution is "Normal." Always say it is "Approximately Normal" because we are using an approximation!
Key Takeaway: The sampling distribution of \(\hat{p}\) is our "crystal ball." It tells us how much we should expect our sample results to vary. If our conditions are met, we can use the Normal distribution to calculate exactly how likely or unlikely a sample result is!
Next Chapter Preview: We will use these concepts to build Confidence Intervals, which allow us to estimate the population proportion when it's unknown!