T1.6 Percentiles, box plots, z-scores, and the empirical rule Start ★ 0
T1.6Playlist 1, episode 6

Is 12 minutes slow?

T1.4 found the middle of a crowd and T1.5 measured its spread. Today: where does one value stand in its crowd? Guess first: you will see your answer again later.

Builds on: T1.4, T1.5Leads to: T1.7, T4.2, T5.2, T7.2, T8.5
0Last time: the see-saw and the square

Same average, two weeks

Seven deliveries each week (invented), both balancing at 10 minutes. Yellow band: one standard deviation to each side. Pink: the 12-minute delivery.

Where a value stands depends on the middle and on the spread. Today: two ways to say where it stands.
Instruments: See-saw (T1.4) and the average square (T1.5).

1Where do I stand?

Where do I stand?

Three everyday places where one value is compared with its crowd. Answer the question on each card.

0 of 3
Two ways to say where you stand: count your place in the line, or measure how far you are from the middle.
2Count

Line up, then count

Lina’s last 19 deliveries, in minutes (invented). Line them up from fastest to slowest. Then slide p and count p% of the line from the fast end.

The pth percentile: the value with about p% of the line-up at or below it.
In French: centile
Instrument: Line-up, the dot crowd of T1.2 standing in order.
25%

2Count

Three scissors, four equal groups

Drag the three scissors along the line-up. Each one cuts through a delivery. Make four groups of the same size.

The three cuts are the quartiles: Q1 (25th percentile), Q2 (the median) and Q3 (75th percentile).
In French: quartiles
2Count

Fold the line into a box

Keep five deliveries: the fastest, the three cuts and the slowest. Then let the line fold into one small picture.

Five-number summary: smallest, Q1, median, Q3, largest. The box holds the middle half; its width is the interquartile range, IQR = Q3 − Q1.
In French: boîte à moustaches · écart interquartile

2Count

When does a dot pop out?

Two dashed fences stand 1.5 box-widths outside the box. Drag the slowest delivery to the right, past the fence.

Fences: Q1 − 1.5 × IQR and Q3 + 1.5 × IQR. A value beyond a fence is an outlier, drawn on its own; the whisker stops at the last value inside.
In French: valeur aberrante
2Count

Four days, four boxes

The 244 real restaurant tables from T1.2, split by day. One box plot per day, on the same scale. Stars are outliers.

Box plots side by side compare crowds in one glance: middles, middle halves and outliers.
3Measure

A ruler made of standard deviations

Ruler turns into a stretchy ruler. Drag its 0 onto the balance point, then stretch its first tick to the edge of the yellow band.

z = the number of ticks from the mean to a value: z = (x − x̄) ÷ s.
Instrument: σ-ruler. Back in T1.7, T4.2, T5.2, T7.2 and T8.5.

3Measure

Two tests, two rulers

The third door. You scored 80 in math and 85 in French (invented classes of 11). Answer first, then lay a σ-ruler under each class.

A z-score has no unit, so it compares values from different crowds: the higher z stands higher in its own crowd.
In French: cote z (Québec) · score z

Which score stands higher in its own class?

4Shape rules

Count the ticks

Lay the σ-ruler under three crowds. Light up 1, 2 or 3 ticks on each side of the mean, and count the values inside.

Bell-shaped crowd: about 68%, 95% and 99.7% of values within 1, 2 and 3 ticks. That is the empirical rule.
In French: règle empirique

4Shape rules

Try to beat the floor

Ten dots, any shape you like. The σ-ruler refits itself as you drag. Push dots more than 2 ticks from the mean, where the outside zones turn pink.

For any shape, at least 1 − 1/k² of the values lie within k ticks: 75% within 2, 89% within 3. That is Chebyshev’s theorem.
In French: théorème de Tchebychev
Goal 1: get 2 of the 10 dots beyond 2 ticks.

!The trap

Negative waits?

60 waits on a busy Saturday (invented): most are short, a few are very long. Mean 4.0 minutes, s = 3.9 minutes. Answer before you look.

The 68–95–99.7 rule needs a bell. On a lopsided crowd, check the shape first; only Chebyshev’s floor is safe for every shape.

The bell rule puts about 2.5% of values more than 2 ticks below the mean: under 4.0 − 2 × 3.9 = −3.8 minutes. How many of the 60 waits are there?

5Outliers

Two alarms for one far value

The book gives two outlier alarms: the box plot’s fences, and a z-score beyond ±3. Drag the yellow dot far to the right and watch which alarm rings first.

A far value stretches the very ruler that measures it, so its z grows slowly: with n values it can never pass (n − 1) ÷ √n. The fences stand on the quartiles, which barely move.
Fence alarmz alarm (|z| > 3)
6Name it

Name what you did

Each card is something your hand already did. Tap a card, then its name. The book’s symbols appear as you match.

0 of 8
Count in the line-up: percentiles, quartiles, box plots. Measure with the σ-ruler: z-scores, the empirical rule, Chebyshev.
7Real data

Your table among the 244

Slide to your own tip or bill. The page counts the tables below you, reads your z on the σ-ruler, and checks both alarms.

Count and measure answer different questions: your place in line (percentile) and your distance from the middle in standard deviations (z).
$4.00

8Practice

Practice arcade

Six levels, a boss where you build a box plot from eleven dots, and a lab where you add yourself to any crowd.

9Remember

Six things to remember, and one trap

Sources and textbook links

Percentiles, quartiles, the five-number summary, box plots, z-scores, Chebyshev’s theorem, the empirical rule and the two outlier rules (fences at 1.5 × IQR, and |z| > 3) follow chapter 3 of the course textbook [1]. This page finds the pth percentile with the rule i = (p/100) × n, rounded up when it is not a whole number; newer editions of the same book series place it at (p/100) × (n + 1). Every count-by-hand example on this page was chosen so that both rules give the same answer. Software uses other rules again [3], so a computer may place the real quartiles a few cents differently. The 244 restaurant tables are real data [2]. The 1.5 × IQR fences go back to Tukey’s box plot [4]. The limit (n − 1) ÷ √n on any z-score is from [5] and is noted in [6]. The weeks, deliveries, classes, heights and waits are invented for teaching; the 200 heights were built to follow a bell shape. “In French” labels give common French statistical vocabulary, not quotations from the book.

  1. Anderson, D. R., Sweeney, D. J., Williams, T. A., Camm, J. D., & Cochran, J. J. (2015). Statistiques pour l’économie et la gestion (5e éd., traduction de la 7e édition américaine d’Essentials of Statistics for Business and Economics par C. Borsenberger). De Boeck Supérieur. Chapitre 3.
  2. Bryant, P. G., & Smith, M. (1995). Practical data analysis: Case studies in business statistics. Irwin. Dataset “tips” (244 tables), as distributed with the R package reshape2 and the seaborn-data repository: https://github.com/mwaskom/seaborn-data
  3. Çetinkaya-Rundel, M., & Hardin, J. (2024). Introduction to Modern Statistics (2nd ed.). OpenIntro. Section 5.5, Box plots, quartiles, and the median. https://openintro-ims.netlify.app/explore-numerical
  4. Tukey, J. W. (1977). Exploratory data analysis. Addison-Wesley.
  5. Shiffler, R. E. (1988). Maximum Z scores and outliers. The American Statistician, 42, 79–80.
  6. NIST/SEMATECH. e-Handbook of statistical methods, section 1.3.5.17, Detection of outliers. https://www.itl.nist.gov/div898/handbook/eda/section3/eda35h.htm