High School Statistics — Unit 1: Exploring Data

Complete Study Guide | Lessons 1.1 – 1.7

📊 THE BIG PICTURE — HOW ALL SEVEN LESSONS FIT TOGETHER

Statistics as a subject starts with one core skill: looking at data critically instead of accepting it at face value. Every lesson in this unit adds one more layer of precision to that skill.

THE FULL ARC OF UNIT 1:

1.1                1.2              1.3             1.4
Question    →     Classify    →    Describe    →   Measure
what you see       & compare        the SHAPE       center & spread
(misleading         two            of quantitative   PRECISELY
 graphs)          categorical       data visually    (mean, median,
                   variables        (histograms,      SD, IQR)
                                    skew, dot plots)


1.5                    1.6                1.7
Package it into   →    Locate a value's → Standardize a value
a boxplot,              exact RANK         so you can compare
compare groups           (percentile)       it ACROSS different
using CSOCS                                  datasets (z-score)


THE THROUGHLINE:
You start by learning to DISTRUST a graph at face value (1.1).
Then you learn the TWO fundamental data types and how each is
described (1.2 categorical, 1.3 quantitative).
Then you learn to put EXACT NUMBERS on a quantitative distribution
(1.4 center/spread, 1.5 boxplots).
Then you learn to LOCATE any single value within that distribution —
first by rank (1.6 percentile), then by standardized distance
from the mean (1.7 z-score).

By the end of Unit 1, you can take ANY data point from ANY dataset
and say exactly where it stands — and you can spot when someone
is using data to mislead you.

🎯 THE EXAM BLUEPRINT — YOUR TEACHER'S THREE OBJECTIVES

Your teacher gave you the actual grading lens for this assessment. These three objectives are not extra — they ARE the test. Everything in Lessons 1.1–1.7 exists to serve one of these three skills. Here is exactly how they map.

OBJECTIVE 1: "Summarize, represent, and interpret data on a
              single count or measurement variable."

THIS MEANS: given ONE dataset, can you fully describe it?

SATISFIED BY:
  Lesson 1.1 → correctly classify the variable (categorical/quantitative)
              and represent it without being misled by a bad graph
  Lesson 1.2 → for categorical: frequency tables, bar graphs, pie charts
  Lesson 1.3 → for quantitative: dot plots, stemplots, histograms, shape
  Lesson 1.4 → compute mean, median, mode, range, IQR, standard deviation
  Lesson 1.5 → build a five-number summary and boxplot; full CSOCS
  Lesson 1.6 → locate a specific value using percentile
  Lesson 1.7 → locate a specific value using a z-score

WHAT THIS LOOKS LIKE ON THE TEST:
  "Here is a dataset / graph. Describe it completely."
  → You are expected to compute the right numbers AND describe
    shape, center, spread, and outliers in words — not just one
    or the other.

OBJECTIVE 2: "Use statistics appropriate to the shape of the
              data distribution to compare the center (median
              or mean) and spread (IQR or standard deviation)
              of two or more different data sets."

THIS MEANS: given TWO (or more) datasets, can you choose the
            RIGHT tool for each one, based on ITS shape?

THE CORE RULE THIS OBJECTIVE IS TESTING (from Lesson 1.4/1.5):

  IF a distribution is roughly SYMMETRIC, no major outliers:
    → report MEAN and STANDARD DEVIATION

  IF a distribution is SKEWED or has OUTLIERS:
    → report MEDIAN and IQR

  THIS CHOICE MUST BE MADE SEPARATELY FOR EACH DATASET.
  Two datasets being compared might even require DIFFERENT
  tools if their shapes differ — e.g., Dataset A (symmetric)
  uses mean/SD, while Dataset B (skewed) uses median/IQR.
  You do NOT have to use the same pair of statistics for both
  just because you're comparing them.

WHAT THIS LOOKS LIKE ON THE TEST:
  "Here are two boxplots / histograms / datasets. Compare their
   centers and spreads."
  → Step 1: identify the shape of EACH one first
  → Step 2: choose mean/SD or median/IQR for EACH one based on
    its own shape
  → Step 3: state the actual comparison using comparative
    language ("Dataset A's median (X) is greater than Dataset
    B's median (Y)")

OBJECTIVE 3: "Interpret differences in shape, center, and spread
              in the context of the data sets, accounting for
              possible effects of extreme data points (outliers)."

THIS MEANS: can you explain WHY the numbers differ, in plain
            English, tied to the actual subject of the data —
            and can you flag when an outlier is DISTORTING the
            picture?

THIS IS THE OBJECTIVE MOST STUDENTS LOSE POINTS ON.
It is not enough to say "Dataset A has a higher mean than
Dataset B." You must:
  • Use the CONTEXT (units, what the data represents) — this
    is the "C" in CSOCS, and it's graded, not decorative
  • Explain what a shape/center/spread difference MEANS in
    that context, not just that it exists
  • Explicitly name any outlier and explain HOW it is affecting
    the mean, range, or standard deviation — and why the median/
    IQR might tell a more honest story instead

WHAT THIS LOOKS LIKE ON THE TEST:
  "Explain why Dataset A's mean is so much higher than its
   median" or "Which measure of spread would you trust more
   here, and why?"
  → This is a WRITTEN interpretation question, not a
    calculation question. See the full worked model below.

The Full Worked Model for Objective 3 (the highest-value skill)

This is the exact type of question most likely to appear, modeled at full credit.

Prompt: Two classes took the same 20-question quiz. Class A's scores: 14, 15, 15, 16, 16, 16, 17, 17, 18, 18. Class B's scores: 5, 14, 15, 16, 16, 17, 17, 18, 19, 20. Compare the two classes' quiz performance, accounting for the effect of any outliers.

STEP 1 — Compute the numbers for EACH class first

Class A: Mean = 16.2, Median = 16, Range = 4, IQR = 2
         SD ≈ 1.23
         → Shape: roughly SYMMETRIC, no outliers
         → Appropriate tools: MEAN and STANDARD DEVIATION

Class B: Mean = 15.7, Median = 16.5, Range = 15, IQR = 3
         SD ≈ 4.14
         → Check for outlier: Q1=15, Q3=17.75(approx), IQR≈2.75
           Lower fence ≈ 15 - 1.5(2.75) ≈ 10.9
           5 is BELOW 10.9 → 5 IS AN OUTLIER
         → Shape: roughly symmetric EXCEPT for one low outlier
         → Appropriate tools: MEDIAN and IQR (because of the outlier)

STEP 2 — Write the full CSOCS-style comparison

"Class A's quiz scores (out of 20) were fairly tightly clustered
around 16, with a small standard deviation (≈1.23) and no
outliers, indicating consistent performance across the class.

Class B's median score (16.5) is actually slightly HIGHER than
Class A's median (16), suggesting that a typical student in
Class B performed marginally better than a typical student in
Class A. However, Class B's mean (15.7) is LOWER than Class A's
mean (16.2) — this is because Class B's mean is being pulled
down by one low outlier: a score of 5, which is unusually low
compared to the rest of that class's scores (which otherwise
range from 14 to 20).

Because of this outlier, the MEAN is not a reliable measure of
Class B's typical performance — the MEDIAN (16.5) gives a more
honest picture. Similarly, Class B's range (15) and standard
deviation (≈4.14) are both inflated by the single low score;
the IQR (≈2.75) is a more trustworthy measure of Class B's
typical spread, and it's still noticeably larger than Class A's
IQR (2), suggesting Class B's scores (aside from the outlier)
were somewhat more varied than Class A's.

CONCLUSION: Without accounting for the outlier, it would look
like Class A slightly outperformed Class B on average. But once
the outlier is identified and the appropriate resistant measures
(median, IQR) are used instead, Class B actually shows a
marginally higher typical score, with somewhat more variability
among the rest of its students."

STEP 3 — Notice what earned full credit
  ✓ Computed appropriate statistics separately for each dataset
  ✓ Identified the outlier explicitly, using the 1.5×IQR rule
  ✓ Explained HOW the outlier distorts mean/range/SD specifically
  ✓ Used the CONTEXT (quiz scores out of 20) throughout, not just
    numbers in the abstract
  ✓ Reached a clear, justified CONCLUSION that accounts for
    the outlier rather than ignoring it

🚨 The #1 point-loss pattern on this type of question: stating numbers without ever explaining what they MEAN in context, or noticing an outlier exists but never explaining HOW it distorts a specific statistic. Both the identification AND the explanation are required.

LESSON 1.1 — MISLEADING WITH DATA

Why This Lesson Exists

A graph can be 100% factually accurate and still create a false impression. This lesson trains you to spot the specific techniques used — intentionally or not — to distort how data is perceived.

Categorical vs. Quantitative Data (The Foundation)

Before spotting misleading graphs, you need to classify what type of data you're looking at — this distinction runs through the entire course.

QUANTITATIVE DATA                    CATEGORICAL DATA
──────────────────────               ──────────────────────────
Numerical values                     Categories or group labels
Values have an inherent ORDER        Often NO inherent order
Makes sense to compute an AVERAGE    Does NOT make sense to average

EXAMPLES:                            EXAMPLES:
• Height (72 inches)                 • Eye color (blue, brown, green)
• SAT score (1350)                   • Favorite sports team
• Number of AP classes (4)           • Dominant hand (left/right)
• Blood pressure (120/80)            • Zip code ← tricky! looks numeric
• Yards per catch                      but it's actually a LABEL
• Income ($45,000)                   • Relationship status

The Quick Test: "Does it make sense to average it?"

Does it make sense to find the AVERAGE SAT score? YES → quantitative
Does it make sense to find the AVERAGE zip code?  NO  → categorical
                        (68494 + 68492) / 2 = 68493 means NOTHING)

🚨 EXAM TRAP: Zip code, phone number, jersey number These LOOK like numbers but they're actually categorical — they're labels, not quantities. You'd never average them meaningfully. Always ask: "does averaging this produce something meaningful?"

The Techniques Used to Mislead With Graphs

TECHNIQUE 1: TRUNCATED Y-AXIS (not starting at zero)

  Normal axis (starts at 0):        Truncated axis (starts at 90):
  100│                              100│▓▓
   80│  ▓▓                           95│▓▓  ▓▓
   60│  ▓▓  ▓▓                       90│▓▓  ▓▓
   40│  ▓▓  ▓▓                          └──────
   20│  ▓▓  ▓▓
    0└──────────                     Small real differences look HUGE
  Bars look similar in height        because the axis is zoomed in

WHY IT MISLEADS: A small actual difference (91 vs 96) can be made
to look like one bar is many times taller than another, when the
real percentage difference is tiny.

TECHNIQUE 2: INCONSISTENT SCALE (uneven spacing on an axis)

  Years marked:  2010   2012   2015   2020   2021
  Spacing:       |------|------|------|-|
                 (2 yrs)(3 yrs)(5 yrs)(1 yr)

  All gaps look the SAME width on the graph, but they represent
  different amounts of time. This exaggerates or hides trends.

TECHNIQUE 3: CHERRY-PICKED DATA (showing only part of the story)

  Showing only years where a trend looks good, and leaving out
  years where the trend reverses.

  Example: "Crime dropped this year!" (true)
  But leaving out that crime had been rising for the previous 5 years
  → the single-year drop looks more significant than it is

TECHNIQUE 4: MISLEADING VISUALS (area/size not proportional to value)

  Using images (like people icons, oil barrels, dollar bill stacks)
  where the HEIGHT represents the data value, but the WIDTH also
  changes — making the AREA (and perceived size) misleadingly large.

  Example: If value doubles, but the image is scaled proportionally
  in BOTH height and width, the visual AREA actually quadruples —
  making the increase look far bigger than it is.

TECHNIQUE 5: CHOOSING THE WRONG GRAPH TYPE

  Pie charts should only be used when categories add up to 100%
  and each individual can only belong to ONE category.

  MISUSE: Using a pie chart when percentages add up to MORE than 100%
  (e.g., survey where people could select multiple options)
  → A pie chart implies a whole being divided into parts,
    which becomes meaningless if the parts overlap
  → Should use a BAR GRAPH instead

TECHNIQUE 6: NO CONTEXT / MISSING LABELS

  A graph missing:
  • Axis labels (what does each axis represent?)
  • Units (dollars? percent? thousands?)
  • Source of the data (who collected it? when?)
  • Sample size (n = 10? n = 10,000?)

  Without these, a graph invites misinterpretation

The Checklist for Evaluating Any Graph

BEFORE TRUSTING A GRAPH, ASK:
☐ Does the y-axis start at zero? If not, is that justified?
☐ Is the scale on each axis consistent (equal spacing = equal amounts)?
☐ Does the graph show ALL relevant data, or a cherry-picked slice?
☐ If it's a pie chart, do the categories add up to 100% with no overlap?
☐ Are there clear labels, units, and a data source?
☐ Does the visual size (area) match the actual data proportionally?

Worked Example — Spotting the Trick

"A truck company advertises: 'Our trucks are 3 TIMES more dependable!' with a bar graph showing their truck's bar three times taller than a competitor's."

STEP 1: Check the y-axis — does it start at 0?
  If the axis starts at 80% instead of 0%, and one truck is at 90%
  reliability and another is at 97%, the DIFFERENCE is only 7
  percentage points — but visually, with a truncated axis, the bar
  can be drawn to look 3x taller.

STEP 2: Check the real percentage difference.
  90% vs 97% is NOT "3 times more dependable" — it's about 7.8%
  more dependable in relative terms. The "3 times" claim likely
  comes from visually measuring the bar heights on a truncated
  graph, not the actual data.

CONCLUSION: The graph is technically showing accurate numbers,
but the truncated axis creates a false visual impression of a
dramatic difference that doesn't reflect reality.

✅ Checkpoint — Lesson 1.1

1. Classify each variable as categorical or quantitative: (a) number of siblings, (b) favorite ice cream flavor, (c) jersey number, (d) blood type, (e) time spent on homework (minutes)

2. A graph's y-axis goes from 70 to 100 instead of 0 to 100. What effect does this have on how differences between bars appear?

3. Why is a pie chart inappropriate for a survey where respondents could select multiple answers?

<details> <summary>Click for Answers</summary>

1. (a) Quantitative — makes sense to average number of siblings. (b) Categorical — flavor is a label. (c) Categorical — despite being a number, it's a label/ID, not a quantity. (d) Categorical — blood type is a category. (e) Quantitative — time has an inherent order and can be averaged.

2. Truncating the axis exaggerates the visual difference between bars — small actual differences appear much larger than they really are.

3. Pie charts represent parts of a whole that should add up to 100%. If people can select multiple answers, percentages sum to more than 100%, which a pie chart cannot accurately represent — a bar graph should be used instead.

</details>

LESSON 1.2 — DESCRIBING A CATEGORICAL VARIABLE

The Core Tools for Categorical Data

Categorical data is organized and described using tables and specific types of graphs — never with a mean or median (remember: you can't average categories).

Frequency Tables

A frequency table organizes categorical data by counting how many observations fall into each category.

FREQUENCY = the count (number) of observations in a category
RELATIVE FREQUENCY = the proportion or percentage in a category

Example — Marital Status of 500 Adults:

Category    Frequency    Relative Frequency
Married     260          52%
Single      150          30%
Divorced    60           12%
Widowed     30           6%
TOTAL       500          100%

Relative frequency = frequency ÷ total
260 / 500 = 0.52 = 52%

Bar Graphs vs. Pie Charts

BAR GRAPH                           PIE CHART
─────────────────────               ─────────────────────
Bars separated from each other      Wedges of a circle
Height = frequency or %             Wedge size = proportion
Works even if % don't add to 100%   ONLY works if % add to 100%
Easy to compare category sizes      Good for showing "parts of a whole"
Can be ordered any way              Order doesn't usually matter

When to use which:

Use a BAR GRAPH when:
• Categories don't need to add to 100% (people can pick multiple)
• You want to easily COMPARE sizes between categories
• There's a category like "Other/Unknown" that changes the total

Use a PIE CHART when:
• Every individual belongs to EXACTLY one category
• Percentages add up to exactly 100%
• You want to emphasize "this is a slice of the whole"

Two-Way Tables — Analyzing TWO Categorical Variables

This is the heart of Lesson 1.2. A two-way table (also called a contingency table) displays data for two categorical variables at once, organized by rows and columns.

EXAMPLE: Police Stops by Race and Whether a Weapon Was Found

                Weapon Found    No Weapon Found    TOTAL
White           20              180                 200
Black           15              285                 300
TOTAL           35              465                 500

Marginal Distributions

The marginal distribution of a single variable is found by looking at the ROW TOTALS or COLUMN TOTALS (the "margins" of the table) — it's essentially the one-way table for that single variable, ignoring the other variable.

MARGINAL DISTRIBUTION OF RACE (using row totals):
White: 200/500 = 40%
Black: 300/500 = 60%

MARGINAL DISTRIBUTION OF WEAPON FOUND (using column totals):
Weapon found: 35/500 = 7%
No weapon found: 465/500 = 93%

KEY IDEA: Marginal distributions look at ONE variable by itself,
ignoring the other variable completely — as if the two-way table
were "collapsed" down to a one-way table.

Conditional Distributions — The Most Important Concept

A conditional distribution shows the distribution of one variable, GIVEN a specific value of the other variable ("conditional ON" that value). This is how you actually investigate whether two categorical variables are ASSOCIATED.

CONDITIONAL DISTRIBUTION OF "WEAPON FOUND," GIVEN RACE:

Given White:  20/200 = 10% had a weapon found
Given Black:  15/300 = 5% had a weapon found

COMPARE THESE PERCENTAGES to check for association:
10% vs 5% → these are DIFFERENT
→ This suggests race and weapon-found ARE associated in this data

Association vs. No Association

NO ASSOCIATION (independence):
  Conditional distributions are approximately the SAME
  regardless of which group you condition on.

ASSOCIATION EXISTS:
  Conditional distributions are DIFFERENT across groups.
  (knowing one variable's value changes what you'd expect
   for the other variable)

🚨 EXAM TRAP: Association is NOT the same as causation. Just because two categorical variables show association in a two-way table does NOT mean one causes the other. There could be lurking (confounding) variables driving both.

Segmented (Stacked) Bar Graphs

A segmented bar graph visually displays conditional distributions — each bar represents one group, divided into segments showing the percentage breakdown of the second variable.

VISUAL:
100%│████████████│░░░░░░░░░░░░│
    │ 90% no wpn  │  95% no wpn│
 50%│             │            │
    │████████████│░░░░░░░░░░░░│
    │ 10% weapon  │  5% weapon │
  0%└─────────────┴────────────┘
       White          Black

If the segments are roughly the SAME height across bars →
no association. If they differ noticeably → association exists.

Simpson's Paradox — The Advanced Trap

Simpson's Paradox occurs when an association or trend seen in SEPARATE groups reverses or disappears when the groups are COMBINED.

CLASSIC EXAMPLE:

Hospital A:  Success rate for surgery = 90% (900/1000 cases)
Hospital B:  Success rate for surgery = 95% (95/100 cases)
→ Hospital B looks better overall!

BUT split by patient severity:

              Hospital A          Hospital B
Mild cases:   95% success (of 800) 98% success (of 90)
Severe cases: 60% success (of 200) 60% success (of 10)

Hospital A treats FAR MORE severe cases (200 vs 10).
When you look within each severity level, the hospitals perform
similarly — but Hospital A's OVERALL rate looks worse only because
it takes on many more difficult cases.

LESSON: Always check for a LURKING VARIABLE (like case severity)
before concluding one group is simply "better" than another based
on a combined/aggregate two-way table.

✅ Checkpoint — Lesson 1.2

Use this two-way table for questions 1-3:

1. Find the marginal distribution of "Studied vs. Didn't Study."

2. Find the conditional distribution of "Passed," given "Studied" and given "Didn't Study." Is there an association?

3. What percentage of ALL students passed the class?

<details> <summary>Click for Answers</summary>

1. Studied: 200/300 = 66.7%. Didn't Study: 100/300 = 33.3%.

2. Given Studied: 180/200 = 90% passed. Given Didn't Study: 60/100 = 60% passed. Very different percentages → there IS an association between studying and passing.

3. 240/300 = 80% of all students passed.

</details>

LESSON 1.3 — DESCRIBING A QUANTITATIVE VARIABLE

Displaying Quantitative Data — Dot Plots, Stemplots, Histograms

DOT PLOT: each data point is a dot above a number line
  Good for: small data sets, seeing individual values

    •
    •  •
  • •  •  •
──┼──┼──┼──┼──┼──→
  5  6  7  8  9

STEMPLOT (stem-and-leaf plot): splits each value into a
"stem" (leading digit(s)) and "leaf" (last digit)

  Stem | Leaf
   5   | 2 4 8
   6   | 1 3 3 7
   7   | 0 2 5 5 9
  (Read: 5|2 means 52)
  Good for: seeing exact values AND shape at once

HISTOGRAM: bars show frequency of values within RANGES (bins)
  No gaps between bars (unlike bar graphs for categorical data!)
  Good for: large data sets, seeing overall shape clearly

  Frequency
  10│      ▓▓
   8│      ▓▓  ▓▓
   6│  ▓▓  ▓▓  ▓▓  ▓▓
   4│  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓
   2│  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓
   0└──────────────────────
     0-10 10-20 20-30 30-40 40-50

🚨 EXAM TRAP: Histogram vs. Bar Graph Histograms (quantitative data) have bars that TOUCH — there are no natural gaps because the x-axis is a continuous numerical scale. Bar graphs (categorical data) have bars that are SEPARATED because categories are discrete, unordered groups.

SHAPE — Describing the Distribution's Form

SYMMETRIC / BELL-SHAPED:
Left and right sides are roughly mirror images

        ▄▄
      ▄▄██▄▄
    ▄▄██████▄▄
  ▄▄████████████▄▄
──────────────────────

SKEWED RIGHT (positive skew):
Long tail stretches to the RIGHT
Most data bunched on the LEFT, few large values pull the tail right

  ▄▄
  ██▄▄
  ██████▄▄
  ████████▄▄___▄______
──────────────────────

SKEWED LEFT (negative skew):
Long tail stretches to the LEFT
Most data bunched on the RIGHT, few small values pull the tail left

              ▄▄
            ▄▄██
        ▄▄██████
  ___▄▄████████████
──────────────────────

Memory device: "The skew is named after the direction of the TAIL, not the hump." A long tail pointing right = skewed right, even though most of the data piles up on the left.

NUMBER OF PEAKS:
UNIMODAL: one clear peak (most common)
BIMODAL: two clear peaks (may suggest two different groups mixed together)
MULTIMODAL: more than two peaks
UNIFORM: roughly flat, no clear peak — all values about equally likely

The Critical Connection: Shape Determines Which Measure to Use

SYMMETRIC distribution → MEAN ≈ MEDIAN (they're close together)
SKEWED distribution → MEAN and MEDIAN are PULLED APART

SKEWED RIGHT: MEAN > MEDIAN (large values pull the mean up)
SKEWED LEFT:  MEAN < MEDIAN (small values pull the mean down)

MEMORY DEVICE: "The mean chases the tail."

Worked Example — Moneyball Context

Baseball salaries are a classic right-skewed dataset: most players earn a modest salary, but a small number of superstars earn tens of millions of dollars.

Salaries (in millions): 1, 1.5, 2, 2, 2.5, 3, 3.5, 4, 25

Median: middle value (5th of 9 ordered values) = 2.5 million
Mean: (1+1.5+2+2+2.5+3+3.5+4+25)/9 = 44.5/9 ≈ 4.94 million

NOTICE: Mean ($4.94M) >> Median ($2.5M)
This gap tells you the data is SKEWED RIGHT — the one $25M
superstar salary pulls the mean way up, but doesn't affect
the median at all.

CONCLUSION: For this dataset, the MEDIAN better represents
a "typical" player's salary.

✅ Checkpoint — Lesson 1.3

1. A dataset has mean = 50 and median = 65. What shape would you expect this distribution to have?

2. A histogram shows most values clustered between 60-70, with a long tail of values stretching from 70 up to 100. What is the shape of this distribution?

<details> <summary>Click for Answers</summary>

1. Since mean (50) < median (65), the distribution is likely skewed LEFT (mean gets pulled down below the median by small extreme values).

2. This is skewed right — the bulk of data (60-70) is on the left, with a long tail stretching to the right.

</details>

LESSON 1.4 — MEASURES OF CENTER AND SPREAD

The Salary Example (Core Context)

A company reports 12 employee salaries (in thousands of $):

29, 32, 34, 34, 34, 34, 35, 35, 39, 43, 67, 185

Notice: eleven employees earn between 29K and 43K.
One employee (the owner/CEO) earns 185K.

Measures of Center — Precise Definitions

MEAN (x̄): sum of all values, divided by the count

  x̄ = (Σx) / n

  For the salary data:
  x̄ = 601 / 12 = 50.08 (thousand dollars)

MEDIAN: the middle value when data is sorted
  n = 12 (even) → median = average of 6th and 7th values
  Median = (34 + 35)/2 = 34.5 (thousand dollars)

MODE: most frequent value
  34 appears four times → Mode = 34

The Flaw of Averages — Why Mean ≠ "Typical"

Mean salary: $50,080
Median salary: $34,500

Company spokesperson says: "Our average salary is over $50,000!"

TECHNICALLY TRUE. But MISLEADING — because:
• 11 out of 12 employees (92%) earn LESS than the mean
• The single $185K salary pulls the mean up dramatically
• The median ($34,500) much more accurately represents
  what a "typical" employee actually earns

THIS IS "THE FLAW OF AVERAGES": Reporting only the mean can
create a false impression when a distribution is skewed or
has outliers — because the mean is NOT resistant to extreme values.

🚨 EXAM TRAP: Always ask "mean or median?" when you see the word "average." "Average" is a casual term that could mean either. A rigorous statistician always specifies.

Measures of Spread — Precise Definitions

RANGE = Maximum − Minimum
  Salary range = 185 − 29 = 156 (thousand dollars)
  Uses ONLY 2 values — extremely sensitive to outliers

INTERQUARTILE RANGE (IQR) = Q3 − Q1
  The range of the MIDDLE 50% of the data
  Resistant to outliers

STANDARD DEVIATION (s or σ): the typical distance
  each data point is from the MEAN

  Formula:  s = √[ Σ(xᵢ − x̄)² / (n − 1) ]

  Step by step:
  1. Find the mean (x̄)
  2. Subtract the mean from each value: (xᵢ − x̄)
  3. Square each difference: (xᵢ − x̄)²
  4. Add up all the squared differences: Σ(xᵢ − x̄)²
  5. Divide by (n − 1)
  6. Take the square root

Worked Example — Standard Deviation by Hand

Using a simplified dataset: 4, 6, 8, 10, 12 (mean = 8)

STEP 1: Mean = (4+6+8+10+12)/5 = 40/5 = 8

STEP 2 & 3: Find (xᵢ − x̄)² for each value

x    (x - x̄)    (x - x̄)²
4    4-8=-4      16
6    6-8=-2      4
8    8-8=0       0
10   10-8=2      4
12   12-8=4      16
                 ─────
SUM:             40

STEP 4: Σ(xᵢ - x̄)² = 40
STEP 5: Divide by (n-1) = 40/4 = 10
STEP 6: √10 ≈ 3.16

Standard deviation ≈ 3.16

Interpretation: On average, each data point is about 3.16 units away from the mean of 8.

Comparing the Three Measures of Spread

┌──────────────────┬────────────────────┬────────────────────┐
│  Measure          │  Sensitive to       │  Best used when...  │
│                   │  outliers?          │                     │
├──────────────────┼────────────────────┼────────────────────┤
│  RANGE            │  YES (very)         │  Quick, rough check │
│  IQR              │  NO (resistant)     │  Data is skewed or  │
│                   │                     │  has outliers       │
│  STANDARD DEV.    │  YES                │  Data is roughly    │
│                   │                     │  symmetric/bell-    │
│                   │                     │  shaped, no extreme │
│                   │                     │  outliers           │
└──────────────────┴────────────────────┴────────────────────┘

Applying this to the salary data:
Standard deviation ≈ 42.7 (inflated by the 185 outlier)
IQR: Q1 = 34, Q3 = 41 → IQR = 7 (much smaller, more "typical")

The IQR (7) best represents the TYPICAL distance between salaries
because it isn't distorted by the single extreme outlier.

✅ Checkpoint — Lesson 1.4

1. A dataset is: 5, 7, 7, 8, 9, 50. Find the mean and median. Which measure is more representative of a "typical" value, and why?

2. Calculate the standard deviation of: 2, 4, 6, 8, 10 (mean = 6). Show all steps.

<details> <summary>Click for Answers</summary>

1. Mean = 86/6 ≈ 14.3. Median = (7+8)/2 = 7.5. The median is more representative — the mean is pulled way up by the outlier value 50.

2.

x    (x-6)   (x-6)²
2    -4      16
4    -2      4
6    0       0
8    2       4
10   4       16
             ──
SUM:         40

s = √(40/4) = √10 ≈ 3.16

</details>

LESSON 1.5 — BOXPLOTS, CSOCS, AND COMPARING DISTRIBUTIONS

The CSOCS Framework — Your Official Description Checklist

Every time you're asked to "describe the distribution," run through these five letters:

C — CONTEXT     What is this data about? (units, subject, source)
S — SHAPE        Symmetric? Skewed left/right? Uni/bimodal?
O — OUTLIERS     Are there any unusual values? (use 1.5×IQR rule)
C — CENTER       Mean or median — report the appropriate one
S — SPREAD       Range, IQR, or standard deviation — report the
                  appropriate one

MEMORY DEVICE: "Careful Statisticians Observe Careful Spreads"

🚨 Simply saying "the data is skewed right" is incomplete. A full CSOCS response mentions all five elements, in context, with units.

Building the Boxplot

FIVE-NUMBER SUMMARY: Minimum, Q1, Median, Q3, Maximum

HOW TO FIND Q1 AND Q3:
1. Order the data from smallest to largest
2. Find the median (splits data into two halves)
3. Q1 = median of the LOWER half
4. Q3 = median of the UPPER half

EXAMPLE: 29,32,34,34,34,34,35,35,39,43,67,185  (n=12, even)

Lower half (first 6): 29,32,34,34,34,34  → Q1 = (34+34)/2 = 34
Upper half (last 6):  35,35,39,43,67,185 → Q3 = (39+43)/2 = 41

Five-number summary: Min=29, Q1=34, Median=34.5, Q3=41, Max=185

Drawing and Reading a Boxplot

        outlier
           |
Min────────┤═══════╪═══════├──────────────────────────Max(outlier)
           Q1    Median   Q3

     └──────┴───────┴───────┘
          the "box" = IQR (middle 50% of data)

Percentage Breakdown Within a Boxplot — Very Testable

  Min ─────── Q1 ─────── Median ─────── Q3 ─────── Max
      25%          25%            25%          25%

• Below Q1: 25%          • Above Q1: 75%
• Below the median: 50%  • Above the median: 50%
• Below Q3: 75%          • Above Q3: 25%
• WITHIN the IQR: 50% (the middle half)

Reading Skew From a Boxplot

SKEWED RIGHT boxplot:              SKEWED LEFT boxplot:
├──┤════╪══════├────────┤          ├────────┤══════╪════┤──┤
Min Q1 Med    Q3        Max        Min      Q1    Med  Q3 Max
Box compressed toward LEFT         Box compressed toward RIGHT
Long tail on the RIGHT             Long tail on the LEFT

SYMMETRIC boxplot:
├────┤════╪════├────┤
Min  Q1  Med   Q3   Max
Whiskers roughly equal, median centered in the box

Identifying Outliers on a Boxplot (Review + Extension)

Using the salary example: Q1=34, Q3=41, IQR=7

Lower fence = Q1 − 1.5(IQR) = 34 − 10.5 = 23.5
Upper fence = Q3 + 1.5(IQR) = 41 + 10.5 = 51.5

Any value below 23.5 or above 51.5 is an outlier.
67 and 185 are BOTH above 51.5 → both are outliers.

Comparing Two or More Distributions

RULE 1: Use COMPARATIVE language for every feature
  NOT: "Group A has a mean of 50. Group B has a mean of 65."
  YES: "Group A's mean (50) is LESS than Group B's mean (65)."

RULE 2: Compare EVERY part of CSOCS, not just one piece

Worked Example — The "Equal Averages" Trap

Two tax cut plans (A and B) both report an average tax cut of $1,260. A politician claims "both plans are identical."

THE TRAP: Equal MEANS do NOT mean equal distributions.

Plan A: mostly small cuts ($200-$400) for most people, with a
        few enormous cuts ($20,000+) for the wealthiest taxpayers
        pulling the mean up to $1,260 → SKEWED RIGHT

Plan B: fairly EQUAL cuts of around $1,200-$1,300 for almost
        everyone → SYMMETRIC

Same mean. COMPLETELY different real-world impact.

✅ Checkpoint — Lesson 1.5

1. A boxplot has Min=10, Q1=25, Median=30, Q3=32, Max=90, with 90 marked as an outlier. Describe the shape.

2. What percentage of data lies between Q1 and the Median?

<details> <summary>Click for Answers</summary>

1. Skewed right — the box is compressed on the left (Q1=25 to Q3=32 is narrow), with a long stretch from Q3 to the outlier at 90.

2. 25% — each quartile section represents exactly 25% of the data.

</details>

LESSON 1.6 — PERCENTILES

What Is a Percentile?

A percentile tells you what percentage of the data falls AT OR BELOW a given value.

DEFINITION:
The p-th percentile of a dataset is the value below which
p% of the data falls.

EXAMPLE:
If you scored in the 85th percentile on a test:
→ 85% of test-takers scored AT OR BELOW your score
→ Only 15% of test-takers scored HIGHER than you

IMPORTANT: Percentile is about RANK/POSITION, not the percentage
of questions you got right!

Percentiles You Already Know

Q1 = the 25th percentile
Median = the 50th percentile
Q3 = the 75th percentile

Calculating a Percentile Rank

Percentile Rank = (Number of values BELOW x) / (Total number of values) × 100

EXAMPLE:
Dataset: 12, 15, 18, 20, 22, 25, 28, 30, 33, 40  (n=10)
Find the percentile rank of 25:
Values below 25: 12, 15, 18, 20, 22 → 5 values
Percentile Rank = 5/10 × 100 = 50th percentile

🚨 EXAM TRAP: Percentile ≠ Percentage correct/score A 90th percentile on a test does NOT mean you scored 90%. It means you scored better than 90% of other test-takers.

🚨 EXAM TRAP: Higher percentile is not always "better" For race times: a LOWER time is better, so being in the 10th percentile (fastest 10%) could be the BEST outcome, not the worst. Always think about what the variable MEANS.

✅ Checkpoint — Lesson 1.6

1. In a class of 30 students, a student's test score places them ahead of 24 other students. What percentile are they in?

2. Dataset: 5, 8, 10, 12, 15, 18, 20, 22, 25 (n=9). Find the percentile rank of the value 15.

<details> <summary>Click for Answers</summary>

1. 24/30 × 100 = 80th percentile.

2. Values below 15: 5, 8, 10, 12 → 4 values. Percentile rank = 4/9 × 100 ≈ 44.4th percentile.

</details>

LESSON 1.7 — Z-SCORES (STANDARDIZED VALUES)

Why We Need Z-Scores

Percentiles tell you rank, but not HOW FAR above or below the center a value is. Z-scores measure exactly how many standard deviations a value is from the mean.

The Z-Score Formula

z = (x − x̄) / s

x = the specific data value
x̄ = the mean of the dataset
s = the standard deviation of the dataset

IN WORDS: "How many standard deviations is this value
           away from the mean?"

Interpreting a Z-Score

z = 0     → the value IS the mean (exactly average)
z = 1     → 1 standard deviation ABOVE the mean
z = -1    → 1 standard deviation BELOW the mean
z = 2     → 2 standard deviations ABOVE the mean

POSITIVE z-score → above the mean
NEGATIVE z-score → below the mean
LARGER |z| → further from the mean (more unusual)

Worked Example — Basic Z-Score Calculation

A class's test scores have mean = 75 and standard deviation = 8. A student scores 91.

z = (91 − 75) / 8 = 16/8 = 2.0

This student scored 2 standard deviations ABOVE the mean —
a notably strong performance.

The Real Power of Z-Scores — Comparing Across Different Distributions

CLASSIC EXAMPLE — "LeBron or Jordan?"

How do you compare two basketball players from DIFFERENT ERAS,
where scoring averages and league conditions were different?

Player A: era mean = 22 ppg, era SD = 4 ppg, scores 30 ppg
  z = (30-22)/4 = 2.0

Player B: era mean = 18 ppg, era SD = 3 ppg, scores 27 ppg
  z = (27-18)/3 = 3.0

Even though Player B's raw average (27) is LOWER than Player
A's (30), Player B's z-score (3.0) is HIGHER — meaning Player B
was MORE exceptional relative to their own era.

Worked Example — Comparing a Test Score and an SAT Score

A student scores 85 on a class test (mean=75, SD=5). Same student scores 1300 on the SAT (mean=1050, SD=200).

Class test z-score: (85-75)/5 = 2.0
SAT z-score: (1300-1050)/200 = 1.25

The student performed BETTER, relative to their peer group,
on the CLASS TEST (z=2.0) than on the SAT (z=1.25) — even
though the SAT score sounds more impressive as a raw number.

Z-Scores and Percentiles — The Empirical Rule

For roughly NORMAL (bell-shaped, symmetric) distributions:

     -3    -2    -1     0     1     2     3      ← z-score
      │     │     │  MEAN  │     │     │
      └─────┴─────┴───┬───┴─────┴─────┘
            68% of data within z = -1 to +1
      ├───────────────────────────────────┤
            95% of data within z = -2 to +2
      ├─────────────────────────────────────────┤
            99.7% of data within z = -3 to +3

QUICK REFERENCE:
z = 0    →  ~50th percentile
z = 1    →  ~84th percentile
z = -1   →  ~16th percentile
z = 2    →  ~97.5th percentile
z = -2   →  ~2.5th percentile

🚨 EXAM TRAP: The Empirical Rule ONLY applies to roughly symmetric, bell-shaped (Normal) distributions. If a distribution is heavily skewed (like the salary data from 1.4), you CANNOT use the 68-95-99.7 rule.

✅ Checkpoint — Lesson 1.7

1. A dataset has mean = 100, SD = 15. Find the z-score for a value of 130.

2. Two students take different exams. Student A scores 78 on Exam 1 (mean=70, SD=4). Student B scores 85 on Exam 2 (mean=80, SD=10). Who performed better relative to their exam's distribution?

<details> <summary>Click for Answers</summary>

1. z = (130-100)/15 = 2.0

2. Student A: z = (78-70)/4 = 2.0. Student B: z = (85-80)/10 = 0.5. Student A performed better relative to their distribution.

</details>

🚨 THE MASTER EXAM TRAP COLLECTION — ALL 7 LESSONS

Lesson 1.1

• Zip codes, phone numbers, jersey numbers are categorical, not quantitative — ask "does averaging make sense?"

• Truncated axes exaggerate differences — always check if a bar graph's y-axis starts at zero

• Pie charts require categories to sum to 100% with no overlap

Lesson 1.2

• Marginal ≠ Conditional distribution — marginal is one variable alone; conditional is given a value of the other

• Association ≠ Causation

• Simpson's Paradox — an aggregated table can show a different (even reversed) trend than subgroups

Lesson 1.3

• Skew is named for the TAIL, not the hump — the bulk of data is on the OPPOSITE side of the skew's name

• Skewed right → mean > median; Skewed left → mean < median

• Histograms have NO gaps; bar graphs (categorical) DO have gaps

Lesson 1.4

• "Average" is ambiguous — always clarify mean or median

• Standard deviation is inflated by outliers — for skewed data, use IQR instead

• Range and standard deviation are NOT resistant to outliers; IQR and median ARE

Lesson 1.5

• Equal means/medians do NOT imply equal or similar distributions

• Each boxplot quarter represents exactly 25% of the data — regardless of its visual width

• When comparing distributions, cover ALL of CSOCS, not just one piece

Lesson 1.6

• Percentile ≠ percentage score — 90th percentile does NOT mean 90% correct

• Higher percentile isn't automatically "better" — depends on what's being measured (e.g., race time)

Lesson 1.7

• Z-score sign matters — positive is above the mean, negative is below

• The 68-95-99.7 Empirical Rule ONLY works for symmetric, bell-shaped distributions — never apply it to skewed data

• Z-scores let you compare ACROSS different distributions/scales — raw numbers alone cannot do this fairly

🔗 MIXED PRACTICE — QUESTIONS THAT COMBINE MULTIPLE LESSONS

Real tests rarely isolate one lesson at a time. These problems blend concepts the way yours likely will.

Mixed Problem 1 (combines 1.1 + 1.3)

A news article shows a bar graph with the y-axis starting at 50 (instead of 0), claiming "test scores have skyrocketed!" The underlying data is: 68, 71, 72, 74, 75, 75, 76, 78, 98.

(a) What misleading technique is being used in the graph? (b) Find the mean and median of the actual data. What shape is this distribution, and how do you know?

<details> <summary>Answer</summary>

(a) Truncated y-axis (starting at 50 instead of 0) — this exaggerates the visual differences between scores, making modest score increases look dramatic.

(b) Mean = (68+71+72+74+75+75+76+78+98)/9 = 687/9 ≈ 76.3. Median = 75 (the 5th of 9 ordered values). Since mean (76.3) > median (75), and there's one clear high outlier (98), the distribution is skewed right.

</details>

Mixed Problem 2 (combines 1.4 + 1.5)

A dataset of home prices (in thousands): 180, 195, 200, 205, 210, 215, 220, 650

(a) Find the five-number summary. (b) Is 650 an outlier? Show your work using the 1.5×IQR rule. (c) Using CSOCS, fully describe this distribution.

<details> <summary>Answer</summary>

(a) Sorted: 180,195,200,205,210,215,220,650 (n=8) Min=180, Max=650 Lower half: 180,195,200,205 → Q1=(195+200)/2=197.5 Upper half: 210,215,220,650 → Q3=(215+220)/2=217.5 Median = (205+210)/2 = 207.5 Five-number summary: Min=180, Q1=197.5, Median=207.5, Q3=217.5, Max=650

(b) IQR = 217.5-197.5 = 20. Upper fence = 217.5+1.5(20) = 247.5. Since 650 > 247.5, 650 IS an outlier.

(c) Context: home prices in thousands of dollars for 8 homes. Shape: skewed right (one extreme high value, most data clustered between 180-220). Outliers: 650 is an outlier by the 1.5×IQR rule. Center: median = 207.5 (thousand) is more appropriate than the mean here due to skew/outlier. Spread: IQR = 20 (thousand) is a more honest measure of typical spread than range (470) or standard deviation, both distorted by the 650 outlier.

</details>

Mixed Problem 3 (combines 1.6 + 1.7)

A student scores in the 84th percentile on a Normally distributed exam with mean=70, SD=6.

(a) Using the Empirical Rule, approximately what z-score corresponds to the 84th percentile? (b) Use that z-score to estimate the student's actual exam score.

<details> <summary>Answer</summary>

(a) From the Empirical Rule quick reference: 84th percentile corresponds to approximately z = 1.

(b) z = (x - x̄)/s → 1 = (x - 70)/6 → x = 70 + 6 = 76

</details>

Mixed Problem 4 (combines 1.2 + 1.5, Simpson's Paradox connection)

Two tutoring companies report their students' average score improvement: Company A: mean improvement = 12 points Company B: mean improvement = 12 points

A parent concludes "these companies are equally effective." Using what you know from Lessons 1.4 and 1.5, explain why this conclusion might be wrong, and what additional information (CSOCS elements) would help evaluate the claim properly.

<details> <summary>Answer</summary>

Equal means do NOT guarantee similar distributions (the "equal averages" trap from 1.5). Company A's 12-point average could come from most students improving by a similar, moderate amount (a symmetric, tightly-clustered distribution) — while Company B's 12-point average could be driven by a few students improving dramatically (30+ points) while most students barely improve at all (a right-skewed distribution with outliers). To properly evaluate the claim, you'd want the full CSOCS description for each company: the shape of each distribution, whether outliers are present, the median (not just mean) improvement, and the spread (IQR or SD) — a boxplot comparison would reveal whether "12 points" represents a consistent, typical outcome or is being skewed by a few extreme cases.

</details>

📖 QUICK-REFERENCE GLOSSARY — ALL 7 LESSONS

Use this as your final scan the night before the test.

🗺 COMPLETE MENTAL MAP — UNIT 1

╔═══════════════════════════════════════════════════════════════════╗
║          UNIT 1: EXPLORING DATA — ALL SEVEN LESSONS                ║
╠═══════════════════════════════════════════════════════════════════╣
║                                                                     ║
║  1.1 MISLEADING WITH DATA                                           ║
║  Categorical vs Quantitative: "does averaging make sense?"          ║
║  Tricks: truncated axis, inconsistent scale, cherry-picking,        ║
║  misleading visuals (area), wrong graph type, missing labels        ║
║                                                                     ║
║  1.2 DESCRIBING CATEGORICAL DATA                                    ║
║  Frequency table, bar graph vs pie chart, two-way table             ║
║  Marginal (one var alone) vs Conditional (one var GIVEN other)      ║
║  Association = conditional distributions differ; Simpson's Paradox  ║
║                                                                     ║
║  1.3 DESCRIBING QUANTITATIVE DATA                                   ║
║  Dot plot, stemplot, histogram (no gaps!)                           ║
║  Shape: symmetric / skewed right / skewed left; uni/bimodal         ║
║  Skew right → mean > median   Skew left → mean < median             ║
║                                                                     ║
║  1.4 MEASURES OF CENTER & SPREAD                                    ║
║  Mean (sensitive) vs Median (resistant) — "flaw of averages"        ║
║  Range vs IQR (resistant) vs Standard Deviation                     ║
║  SD formula: s = √[Σ(x-x̄)²/(n-1)]                                ║
║                                                                     ║
║  1.5 BOXPLOTS & CSOCS                                                ║
║  CSOCS: Context, Shape, Outliers, Center, Spread                    ║
║  5-number summary → boxplot (each quarter = 25% of data)            ║
║  Outliers: beyond Q1-1.5(IQR) or Q3+1.5(IQR)                       ║
║  Comparing groups: comparative language on ALL of CSOCS             ║
║                                                                     ║
║  1.6 PERCENTILES                                                      ║
║  p-th percentile = % of data AT OR BELOW that value                 ║
║  Percentile rank = (# below x)/(total) × 100                        ║
║  NOT the same as % score/% correct                                  ║
║                                                                     ║
║  1.7 Z-SCORES                                                         ║
║  z = (x - x̄)/s  → "how many SDs from the mean?"                   ║
║  Lets you compare values ACROSS different distributions             ║
║  68-95-99.7 Rule: ONLY for symmetric/Normal distributions           ║
╚═══════════════════════════════════════════════════════════════════╝

🏁 FINAL REVIEW — FULL UNIT SPEED ROUND

<details> <summary>Click for Answers</summary>

</details>

🎯 One Last Check Against Your Teacher's Three Objectives

• [ ] I can fully describe ONE dataset — right graph, right numbers, right words (Objective 1)

• [ ] I can look at TWO datasets and pick the right tool (mean/SD vs. median/IQR) for EACH one based on ITS OWN shape (Objective 2)

• [ ] I can write a full paragraph explaining WHY two datasets differ, naming any outlier by value and explaining specifically how it distorts which statistic (Objective 3)

If any of those three feel shaky, re-read the Full Worked Model above — that single example is built to hit all three objectives at once.

📋 HOW YOUR ANSWERS ARE ACTUALLY SCORED — THE OFFICIAL RUBRIC PATTERNS

This section is built directly from an official Skew The Script scoring rubric for a Unit 1 "Exploring Data" assessment (86 points across 6 problems, with contexts like social media usage, basketball stats, SAT prep hours, cafeteria satisfaction, video game tournaments, and music streaming). Even without seeing the exact problems, the rubric's own point breakdown reveals the exact scoring patterns your teacher's rubric almost certainly follows too — because it comes from the same source as your curriculum.

The single biggest pattern across all six problems: almost no part awards full credit for a correct number alone. Every category below pairs the calculation with an explanation, comparison, or justification — and partial credit exists specifically for getting the number right but the reasoning weak (or vice versa).

The Recurring Point Structures

PATTERN 1 — Variable classification
  Full credit = ALL variables correctly categorized
  Partial credit = most correct (graded by count, e.g. 3-4 of 5)
  → Sometimes an EXPLANATION for each is required for full credit,
    not just the correct category label

PATTERN 2 — Identifying a misleading graph
  Full credit = naming the misleading FEATURE *and* explaining
  its IMPACT on interpretation, with specific detail
  → Just naming "truncated axis" without explaining what it does
    to how the data looks = partial credit only
  → Some problems require finding MULTIPLE distinct misleading
    features in the same graph (not just one) — read carefully
    for how many the question asks for

PATTERN 3 — Relative frequency / conditional distribution
  Full credit = correct values, in the requested FORMAT
  (some rubrics require BOTH decimal and percentage forms)
  → Conditional distribution questions require the calculation
    AND a clear interpretation AND a meaningful comparison
    between groups for full credit — a bare number is a partial
    answer here, not a complete one

PATTERN 4 — Five-number summary
  Full credit = ALL SIX values: min, Q1, median, Q3, max, AND IQR
  → Note: IQR is graded as part of the "five-number summary"
    question, not a separate follow-up — always compute it
    alongside the five-number summary, don't wait to be asked

PATTERN 5 — Identifying shape
  Full credit = correct shape identified AND justified using
  SPECIFIC boxplot (or histogram) FEATURES
  → "It's skewed right" alone is a partial answer
  → "It's skewed right because the box is compressed toward Q1
    and the whisker/outliers extend much further past Q3" is
    the kind of specific, feature-based justification that
    earns full credit

PATTERN 6 — Choosing mean vs. median, or SD vs. IQR
  Full credit = correct choice, with reasoning that explicitly
  references OUTLIERS or SHAPE
  → Just picking "median" without explaining why (skew/outliers
    present) earns partial credit only

PATTERN 7 — Comparing TWO datasets
  Full credit = compare BOTH measures being asked about
  (e.g., both SD and IQR — not just one), AND compare shape
  of BOTH datasets with reasoning about mean vs. median for each
  → A comparison that only addresses one dataset, or only one
    of the two requested statistics, is a partial answer even
    if that one part is fully correct

PATTERN 8 — Z-scores
  Full credit = correct calculation AND a contextual
  interpretation in words (what the z-score actually MEANS
  for that specific person/value, in that specific context)
  → A bare number with no sentence explaining what it means
    is a partial answer

PATTERN 9 — Working backward from a z-score to a raw value
  Given a z-score and told to find the actual data value it
  corresponds to, full credit requires the correct value AND
  an explanation of how you got there (rearranging z=(x-x̄)/s
  to solve for x) — this is the REVERSE of the usual z-score
  direction, and it's explicitly tested

PATTERN 10 — "Cannot conclude" is sometimes the correct answer
  One rubric awards full credit specifically for concluding that
  you CANNOT determine something definitively, when the evidence
  (e.g., IQR and SD pointing in different directions, or limited
  data) doesn't support a confident claim either way
  → Forcing a confident conclusion when the honest answer is
    "the data doesn't clearly support one conclusion" is
    actually the WRONG answer in this case — recognizing
    genuine ambiguity, backed by reasoning, is what's rewarded

PATTERN 11 — Advantages/disadvantages comparisons
  Some questions ask you to weigh trade-offs directly — e.g.,
  advantages AND disadvantages of a pie chart vs. a bar graph,
  or of one statistical method vs. another
  Full credit = ONE clear advantage AND one clear disadvantage
  (or two of each, depending on the question), each explained
  — not just listed as a word or two

The One-Sentence Summary of This Whole Rubric

A number by itself is never a full-credit answer in this curriculum. Every calculation question is really two questions stacked together: compute it correctly, and explain what it means, why you chose that method, or how it compares — in words, using the context, and (where relevant) naming any outlier or shape feature that affected your reasoning. This is exactly the CSOCS habit and the Objective 3 skill from earlier in this guide — the rubric just confirms, point by point, that it's what's actually being graded.

🚨 If you're ever unsure how much to write: default to including the interpretation sentence. Every single pattern above shows that skipping it costs points, even when the underlying number is correct.

🎯 What's Your Weak Spot?

• [ ] Classifying tricky variables (zip codes, jersey numbers) as categorical

• [ ] Identifying specific misleading graph techniques

• [ ] Building and reading a two-way table; marginal vs. conditional

• [ ] Determining association and explaining Simpson's Paradox

• [ ] Identifying shape (skew direction) from a histogram or boxplot

• [ ] Calculating standard deviation by hand, step by step

• [ ] Explaining the "flaw of averages"

• [ ] Choosing IQR vs. standard deviation based on shape/outliers

• [ ] Reciting and applying CSOCS completely (all 5 letters)

• [ ] Finding Q1 and Q3 correctly; identifying outliers with the 1.5×IQR rule

• [ ] Comparing two distributions using comparative language

• [ ] Distinguishing percentile from percentage score

• [ ] Calculating and interpreting z-scores; comparing across distributions

• [ ] Applying the 68-95-99.7 rule correctly (and knowing when NOT to)

💡 A FEW IDEAS TO GET EVEN MORE OUT OF THIS GUIDE

A few things worth trying as you review, beyond just re-reading:

1. Teach CSOCS out loud, from memory, using a dataset you make up on the spot. Being able to generate your own example (not just recognize one) is a much stronger test of whether you actually know it.

2. For every "skewed right vs skewed left" moment, physically point to where the tail is before answering. This single habit eliminates the most common mistake in the whole unit.

3. Before your test, try writing ONE full CSOCS description from scratch, timed, without looking at this guide. If you can do that cleanly in under 3 minutes, you're in strong shape.

4. If your teacher allows a notecard, the Quick-Reference Glossary above is designed to fit that purpose almost as-is — it's already organized as a compressed, complete reference.

Good luck — this is a genuinely complete foundation for the rest of your stats course. 📊