Topic 6.1 Introducing Statistics: Why Be Normal?: explain how the approximately normal sampling distribution of a sample proportion lets us quantify uncertainty and make inferences about an unknown population proportion.

What this topic is asking

The College Board (Topic 6.1) opens Unit 6 with the idea behind inference: because the sampling distribution of a sample proportion $\hat{p}$ is approximately normal under known conditions, a single sample lets us make a quantified statement about the unknown population proportion $p$. This topic is conceptual; it sets up the confidence intervals and significance tests that fill the rest of the unit.

From one sample to a statement about the population

You never see $p$ directly. You see one $\hat{p}$, which would change if you took a different sample. The breakthrough idea, built across Unit 5, is that this variation is not chaotic: the collection of all possible $\hat{p}$ values forms a sampling distribution with a predictable center, spread, and shape. Topic 6.1 turns that fact into a tool. Because the sampling distribution is centered at $p$, your single $\hat{p}$ is a sensible estimate; because its spread is known, you can say how far off it is likely to be; because its shape is approximately normal, you can convert that into exact percentages.

Why normality is the key

Without an approximately normal sampling distribution, a single $\hat{p}$ would be just a number with no attached uncertainty. Normality supplies the ruler: the $68$-$95$-$99.7$ pattern and z-scores translate "how many standard deviations is $\hat{p}$ from a value of $p$" into a probability. That is why every proportion procedure in Unit 6 begins by checking the large-counts and $10\%$ conditions; they are exactly the conditions that earn the normal model.

The two questions inference answers

Inference about $p$ takes two complementary forms, both powered by the normal sampling distribution.

1. Estimation (confidence intervals). We have no claimed value of $p$; we want a plausible range. Center an interval at $\hat{p}$ and extend it by a margin of error built from the normal model: $\hat{p} \pm z^{*}\sqrt{\hat{p}(1-\hat{p})/n}$.
2. Testing (significance tests). Someone claims a specific $p$ (for example $p = 0.5$). We ask how surprising our $\hat{p}$ would be if that claim were true, using a z-statistic and a P-value.

Both rest on the same picture: $\hat{p}$ is one point on a known bell curve. Recognizing that a single statistic is a draw from a distribution, not the truth itself, is the central reasoning move of the entire inference half of the course, and it is why Topic 6.1 is framed as an idea rather than a calculation.

A caution built into the idea

Inference quantifies uncertainty; it does not remove it. A confidence interval can miss $p$, and a test can reach the wrong conclusion, because $\hat{p}$ genuinely varies. The normal model tells you how often that happens (the confidence level, the error rates of Topic 6.7), which is the honest alternative to pretending one sample reveals the truth exactly. This is the mindset every later topic depends on.

Try this

Q1. State the parameter and the statistic in a study estimating the proportion of voters who support a measure. [1 point]

• Cue. Parameter: $p$, the true proportion of all voters who support it. Statistic: $\hat{p}$, the sample proportion who support it.

Q2. Why does proportion inference require the large-counts condition? [1 point]

• Cue. It is what makes the sampling distribution of $\hat{p}$ approximately normal, and normality is what lets us attach probabilities (margins of error, P-values) to a single $\hat{p}$.