🔔 z-Scores & the Normal Curve

1. The big idea — a score with no context is silent

A raw score by itself is almost meaningless. 32 stress points tells you nothing until you ask the only question that matters: compared to whom? Thirty-two is alarming if the class averages 20 and nearly impossible if it averages 60. The number alone can't speak — it needs the mean (where the group centers) and the standard deviation (how spread out the group is) before it means anything at all.

A z-score supplies exactly that context and folds it into a single number. It rescales any raw score into one universal language: how many standard deviations away from the mean it is. Once a score has been turned into a z, it has been standardized — stripped of its original units and placed on a common ruler that every other standardized score shares.

A z of +2 means "two standard deviations above average." A z of −1 means "one standard deviation below." A z of 0 means you are the average. That's the whole idea — the rest of this unit is just learning to compute it, picture it, and read a percentile off of it.

2. The normal curve (the "bell curve")

Before we standardize anything, look at the shape most of our scores actually take. Pour together a big pile of natural measurements — height, reaction time, test scores, stress — and they tend to settle into one familiar form: tall in the middle, tailing off symmetrically on both sides. We call it the normal curve (or, casually, the bell curve). It shows up so often because most outcomes are pushed around by many small, independent causes that mostly cancel out — a deep mathematical result that's well beyond us here, but worth knowing isn't an accident.

Three features define every normal curve:

• It's symmetric — the left half is a mirror image of the right.
• The mean, median, and mode all sit in the same spot: the exact center, where z = 0.
• It's unimodal — one peak in the middle, thinning steadily as you move toward either tail.

The further you move from the center in either direction, the rarer scores become. That's why a z of +2 is impressive and a z of +3 is genuinely unusual — out there, the curve has almost no height left, so almost nobody lands that far out. Drag the sliders below and watch the curve, the score marker, and the percentile move together:

Notice the two controls do different jobs. Sliding the mean shifts the whole curve left or right without changing its shape — it just moves where "average" is. Sliding the SD changes the spread: a wider SD flattens and stretches the curve, a narrow SD makes it tall and tight. The very same raw score becomes more or less unusual depending on how spread out the group around it is — which is precisely why the z-score formula has to divide by the SD.

3. The 68–95–99.7 rule (the empirical rule)

Here's what makes the normal curve so useful: its proportions are fixed. No matter what the mean or SD happen to be, the share of scores within a given number of standard deviations of the mean is always the same. That regularity is called the empirical rule, or just the 68–95–99.7 rule:

So almost everybody — about 997 of every 1,000 people — lands within 3 SDs of the mean. A z beyond ±3 is genuinely exceptional. The static figure below makes the bands concrete; the shaded region marks a single score (here z = +1) and the area below it:

4. The z-score formula

Two small notes on symbols you'll see all over this course. We write μ (mu) for the mean and σ (sigma) for the standard deviation when we're treating the numbers as a whole population; you'll also see the sample versions x̄ and s. The z-score formula works identically either way — plug in whichever center and spread describe your group.

On OUR data — stress (column J), mean = 20.1, SD = 5.9:

• A student with stress 32: z = (32 − 20.1) ÷ 5.9 = +2.02 → just over two SDs above average — very high stress.
• A student with stress 14: z = (14 − 20.1) ÷ 5.9 = −1.03 → about one SD below average — calmer than most.

On gpa (column K), mean = 3.08, SD = 0.44: a 3.9 GPA gives z = (3.9 − 3.08) ÷ 0.44 = +1.87 → almost two SDs above average — well above the pack. Notice the GPA scale (0–4) and the stress scale (roughly 0–40) are wildly different in size, yet both produce a z near +2. That is the magic we'll cash in shortly.

Try it now — the z-Score calculator

Type a raw score, the mean, and the SD — then read the z-score and percentile straight off. Hit Load example to drop in a real stress score from our data, and check it against the worked example above (X = 26, μ = 20, σ = 6).

5. From z-score to percentile (and back)

A percentile says what percent of the group scored below you. Because the normal curve's proportions are fixed, every z-score maps to exactly one percentile — and that mapping is what a z-table (or a spreadsheet) stores. The empirical rule already gives you the famous landmarks by heart:

Where do the in-between numbers come from? Each is just "50% for the lower half, plus (or minus) half of whatever band you've crossed." For z = +1 you add half of 68% (= 34%) to the bottom 50% and get the 84th. For z = −1 you subtract: 50 − 34 = the 16th. So our student with a stress z of +2.02 scored higher than roughly 98% of classmates. That's a precise, honest statement — and it came from one little formula plus the shape of the curve.

6. Why z-scores make different scales comparable

This is the payoff. Raw scores live in their own units — stress points, GPA, SAT points, inches — and you cannot compare across them directly. A 1300 SAT and a 29 ACT are measured on scales that don't even overlap. But standardizing strips the units away. Once both scores become z-scores, they live on the same ruler, centered at 0 with a standard deviation of 1, and you can lay them side by side. Whoever has the larger z did better relative to their own group — and the comparison is fair even though the original tests share nothing.

The same trick answers a question about a single person across two traits. Is our student more unusual in stress (z = +2.02) or in GPA (z = +1.87)? Ignore the signs and compare the magnitudes: the bigger |z| is the more extreme. Here stress (2.02) edges out GPA (1.87), so this student is a touch more of an outlier in stress than in grades — a sentence you simply could not write from the raw numbers 32 and 3.9.

💻 In Google Sheets

You won't hunt through a paper z-table on exam day — Sheets does both halves for you. Open the SJNY Student Life dataset and try these. First, get the group's center and spread; then standardize any score:

=AVERAGE(J2:J201)                  → the mean (μ) of stress
=STDEV(J2:J201)                    → the standard deviation (σ)
=STANDARDIZE(32, 20.1, 5.9)        → the z-score: (32 − 20.1) ÷ 5.9 ≈ 2.02

=STANDARDIZE(X, mean, SD) is the z formula in one cell — same three inputs, same answer. To turn that z into a percentile (the area of the curve below it), use the standard-normal distribution function:

=NORM.S.DIST(2.02, TRUE)           → 0.978  (about the 98th percentile)
=NORM.S.DIST(1, TRUE)              → 0.8413 (the ~84th percentile from our example)

The TRUE asks for the cumulative area — everything below that z. To run it backwards (percentile → z), use =NORM.S.INV(0.95), which returns about 1.645 — the z that marks the top 5%. Between STANDARDIZE, NORM.S.DIST, and NORM.S.INV, you can move freely among raw scores, z-scores, and percentiles without ever touching a table.

Key terms