
Statistics – Averages and Analysis
MYP Mathematics framework — Statistics strand: averages, spread and grouped-data analysis
Averages tell you almost nothing sitting on their own — a mean of 23 could hide a class where everyone scored around 20, or one where half the class failed and a handful aced it. This chapter is about squeezing the real story out of a set of numbers: grouping raw data sensibly, working backwards from an average to rescue a missing value, reading a box plot in one glance, and — the skill examiners reward most and mark down hardest — actually comparing two datasets instead of just parking two numbers next to each other.
Overview — The Shape of the Chapter
Why grouping, missing values and comparisons all sit in one chapter
- Raw lists of 20–30 numbers are unreadable — grouping into class intervals turns them into a table you can actually interpret at a glance.
- Every 'average' (mean, median, mode) answers a slightly different question about what's typical — mixing them up is the single most common error across this whole unit.
- A statistic means nothing until it's compared: '23 out of 50' sounds mediocre until you learn the other class averaged 15 — or brilliant until you learn they averaged 40.
- Box plots and IQR exist because averages alone hide spread — two classes can share an identical median and still have completely different levels of consistency.
The shape of the chapter
Command terms that actually appear on this topic
| Command term | What it demands | AO | Mark-earning move |
|---|---|---|---|
| State / Identify | Give a specific value or fact straight from the data — no working or justification required. | Knowing | 1 mark for the correct value alone; writing a full interval where a single number was wanted (e.g. '20–29' instead of '24.5') scores 0. |
| Calculate | Produce a numerical answer, showing the method used to get there. | Applying | Method mark for correct working even if the final figure is arithmetically wrong; a bald correct-looking answer with no working can lose the method mark if it's actually wrong. |
| Describe | Give a written account of a pattern or relationship in the data. | Applying | Needs a full sentence about a trend (e.g. 'as width increases, frequency tends to increase') — a single word like 'positive' scores 0. |
| Explain | Give reasons that justify a statement — the 'why', not just the 'what'. | Reasoning | A correct fact with no reasoning attached earns 0 on an Explain mark; the reasoning IS the mark. |
| Compare | Relate two or more data sets directly to each other using a comparative word. | Reasoning | Two separate correct statements about each dataset, with no linking word (higher/lower/more consistent), typically lose the comparison mark entirely. |
Key point
Working with Grouped Data
Building and reading a grouped frequency table
- A class interval (or 'class') is a range of values bundled together in a frequency table so a large dataset becomes readable — '20–29' groups every score from 20 to 29 inclusive into one row instead of listing them individually.
- The class limits are the numbers actually printed in the table — the lower limit (20) and upper limit (29). These are NOT automatically the true dividing points between classes.
- The midpoint stands in for every value inside that class once you've lost the raw data: midpoint = (lower limit + upper limit) ÷ 2.
- The class boundary is the true halfway dividing line between adjacent classes — e.g. between the top of '10–19' (19) and the bottom of '20–29' (20), the boundary sits at 19.5. Boundaries matter because limits like 19 and 20 look like they touch but leave an artificial 'gap' on a continuous number scale.
- Class width = upper boundary − lower boundary. For '20–29': 29.5 − 19.5 = 10.
- The modal class is simply the interval with the highest frequency — you can name the class but you can never state an exact 'mode' once data has been grouped, because you no longer know the individual values.
Beyond the syllabus — Frequency Density. When class widths are NOT equal (like the age groups 0–5, 6–10, 11–20, 21–30), comparing raw bar heights on a histogram is unfair, because a wider bar can rack up a bigger frequency just by covering more ground. Full statistics courses fix this with frequency density = frequency ÷ class width. This isn't required at MYP Year 2 — you're only expected to notice and describe the problem (see Q8 below), not calculate density.
Vocabulary examiners expect you to use precisely
- Class interval
- A range of values grouped together in a frequency table to organise data.e.g. '30–39' for test scores
- Class limit
- The lower or upper number actually written in the table for a class.For '20–29': lower limit 20, upper limit 29
- Class boundary
- The true dividing point between two adjacent classes, found as the midpoint between the top of one interval and the bottom of the next.Between 19 and 20 → boundary 19.5
- Class width
- The size of an interval, found as upper boundary minus lower boundary.29.5 − 19.5 = 10
- Midpoint
- The average of a class's lower and upper limit, used to represent every value inside that class.(20+29)/2 = 24.5
- Modal class
- The class interval with the highest frequency in a grouped table.Not the same as 'the mode' — no individual value is identified
Grouped-data toolkit
average the two stated limits
always use boundaries, not raw limits, for width
estimated mean of grouped data
Histogram of grouped test scores with midpoints marked
Frequencies 2, 5, 8, 4, 1 for classes 0–9 to 40–49 (n = 20) — the data used in the worked mean estimate below.
Worked example
Building a frequency table from a raw list (tallying)
Worked example
Estimating the mean of grouped data and comparing two classes
Worked example
Class boundaries and class width
Worked example
Choosing a sensible number of class intervals
Worked example
Unequal class widths — reading the age-group table
Worked example
Grouped mean vs the exact mean — how good is the estimate?
Limits copied in as boundaries
Writing the printed class limits (e.g. 20 and 29) directly as the 'boundaries' instead of computing the midpoints between adjacent classes.
Consequence: Loses both marks on a 2-mark 'identify the boundaries' question, even though the class interval itself was read correctly.
Fix: Always find the midpoint between the TOP of the interval below and the BOTTOM of the interval in question (and the same going up) — boundaries almost always end in .5 for whole-number data.
'State the midpoint' is a 1-mark, single-number question
Don't forget the final division in a grouped mean
Key takeaways
- Class interval = the group; class limit = the printed numbers; class boundary = the true dividing line (usually .5); class width = boundary difference.
- Midpoint stands in for every value in a class — estimated mean = Σ(midpoint × frequency) ÷ Σfrequency.
- Modal class is the highest-frequency interval — you cannot state an exact mode once data is grouped.
- Too many intervals fragments the data (ties, noise); too few hides the pattern entirely — a moderate number of intervals usually shows the clearest trend.
- Unequal class widths make raw frequency comparisons misleading — width doesn't proportionally determine frequency.
Finding Missing Data with Given Averages
Reverse-engineering the mean
- A mean secretly encodes a total: if the mean of 6 tests is 15, the total of all 6 marks MUST be 15 × 6 = 90, no matter what the individual scores are.
- The method is always the same three moves: (1) work out how many values there should be in total (n), (2) multiply n by the mean to get the required total, (3) subtract the sum of everything you already know — whatever's left belongs to the missing value.
- The exact same algebra handles 'find the missing frequency' or 'find x so that the mean equals ___' questions — only the labels change, the moves don't.
Missing-value algebra
rearrange the mean formula to get the required total
Worked example
Finding a missing test score
Using the wrong n
Multiplying the mean by the number of KNOWN values (5) instead of the TOTAL number of values including the unknown (6) — i.e. computing 15 × 5 = 75 instead of 15 × 6 = 90.
Consequence: Produces a wrong missing value that still looks plausible, so it's rarely caught by a sanity check.
Fix: Before multiplying, count how many data points there are supposed to be IN TOTAL once the missing one is included — that's your n, always.
Show the multiplication step explicitly
Key takeaways
- A mean hides a total: total = mean × n.
- Missing value = required total − sum of known values.
- Always use the FULL n (including the unknown item you're solving for), never just the count of known values.
Box Plots and Interquartile Range
Reading and building a box plot
- A box plot is just a picture of the five-number summary: minimum, lower quartile (Q1), median, upper quartile (Q3), maximum.
- To find quartiles by hand: order the data, find the overall median first, then treat the lower half and upper half as two mini-datasets and find EACH of their medians — that gives Q1 and Q3.
- With an even number of values, the two halves split cleanly with no value shared between them. Always double-check each half has exactly n/2 values before finding its median.
- Reading left to right: whisker (min → Q1), box (Q1 → Q3 with the median line inside), whisker (Q3 → max). The box itself is the middle 50% of the data — the most tightly packed, 'typical' range.
- IQR = Q3 − Q1. It describes how spread out the MIDDLE half of the data is, deliberately ignoring the extremes — which makes it far less sensitive to a single outlier than the range.
Box plot vocabulary
- Quartile
- One of the three values that split ordered data into four equal-sized groups.Q1 (25% mark), median (50% mark), Q3 (75% mark)
- Interquartile range (IQR)
- The difference between the upper and lower quartiles, Q3 − Q1, measuring the spread of the middle 50% of data.Ignores the most extreme values on either end
- Whisker
- The line extending from the box out to the minimum or maximum value.Shows the full range beyond the middle 50%
Quartiles and spread
Box plot of a single dataset
Five-number summary: min 5, Q1 8.5, median 12.5, Q3 17, max 25 — from the worked example below.
Worked example
Finding the five-number summary and IQR
Range vs Interquartile Range
| Concept A | Concept B | The distinction that matters |
|---|---|---|
| Range | Interquartile Range (IQR) | Range uses only the max and min (whole spread, wrecked by a single outlier); IQR uses Q3 − Q1 (middle 50%, barely affected by extreme values) — always report BOTH if a question asks about 'spread' without specifying which. |
Finding quartiles from unordered data
Splitting the data into halves before sorting it into ascending order.
Consequence: Produces a box plot that doesn't match the actual data — examiners can't award method marks if the working shown doesn't correspond to ordered halves.
Fix: Order the full list first, always, before doing anything else with quartiles.
IQR wants ONE number, not two
Key takeaways
- Box plot = picture of min, Q1, median, Q3, max.
- Find quartiles by ordering data, finding the overall median, then finding the median of each half.
- IQR = Q3 − Q1, measuring the spread of the middle 50%, far less sensitive to outliers than range.
Calculating Mean, Median, Mode and Range
Raw data gets the exact formulas — no estimating
- Once you have the actual raw list (not a grouped table), use the real formulas — no midpoints, no estimating.
- Median needs the data ORDERED first, every single time: for odd n take the middle value, for even n average the two middle values.
- Mode is whichever value(s) appear most often. A dataset can have no mode (every value different), one mode, or several — 'no mode' is a completely valid final answer, don't force one.
- Range = max − min: the simplest, crudest measure of spread, using only the two extreme data points and ignoring everything between them — exactly why a single outlier can distort it badly.
Raw-data formulas
mean
Worked example
Mean, median, mode and range of the raw score list
Which average, when?
| Measure | Best used when... | Watch out for |
|---|---|---|
| Mean | Data is roughly evenly spread with no extreme outliers | Easily dragged up or down by just one very high or low value |
| Median | Data has outliers or is skewed (e.g. house prices, salaries) | Ignores the actual size of extreme values — only cares about position |
| Mode | You need the single most common category or score (especially non-numeric data) | May not exist, or may not be unique — never force an answer |
| Range | A quick, crude first impression of spread is needed | Uses only 2 data points; one outlier can inflate it hugely — prefer IQR for a fairer picture |
Mean vs Median
| Concept A | Concept B | The distinction that matters |
|---|---|---|
| Mean | Median | Mean uses every value and gets pulled towards outliers; median is the middle value and stays put regardless of how extreme the outliers are — use median to describe 'typical' when a dataset is skewed. |
Remembering what each measure actually tells you
Mode = Most common. Median = Middle. Mean = Maths (add everything, divide by how many).
Three M's, three different questions about 'typical' — reciting this stops you defaulting to 'mean' automatically when a question actually wants the middle value or the most frequent one.
Forgetting to order data, or forcing a mode
Finding the 'middle value' of an unordered list, or inventing a mode when every value is actually different.
Consequence: Wrong median (picks the wrong value entirely), or an invented mode that has no basis in the data — both lose the mark outright.
Fix: Always write the ordered list out first for median questions; check frequencies carefully before answering a mode question — 'no mode' is a legitimate, markable answer.
Even n means you MUST average the two middle values
Key takeaways
- Mean, median and mode each describe a different sense of 'typical' — know which the question is actually asking for.
- Order data before finding a median; average the two middle values if n is even.
- Mode can be none, one, or several values — never force one that isn't there.
- Range = max − min, a crude spread measure heavily distorted by outliers.
Comparing and Interpreting Averages
Turning numbers into a proper comparison
- 'Compare' is a command term, not a suggestion — the answer must directly relate two values to each other ('higher than', 'more consistent than', 'closer together than'), not just report each number separately.
- Full marks on a comparison question almost always need BOTH halves: a comparison of centre (mean/median) AND a comparison of spread (range/IQR) — two datasets can share an identical average and still behave completely differently.
- Percentages sharpen an interpretation: saying '40% of students scored in the modal class 20–29' is far more informative than just naming the class, because it tells the reader how dominant that group really is.
Worked example
Modal class as a percentage — and what it actually means
Worked example
Same median, very different spread
Comparing two box plots with the same median
Team A: min 10, Q1 12, median 15, Q3 18, max 20 (IQR 6). Team B: min 8, Q1 9, median 15, Q3 21, max 22 (IQR 12) — identical medians, very different spread.
Team A vs Team B — full statistical comparison
| Statistic | Team A | Team B |
|---|---|---|
| Minimum | 10 | 8 |
| Lower Quartile (Q1) | 12 | 9 |
| Median | 15 | 15 |
| Upper Quartile (Q3) | 18 | 21 |
| Maximum | 20 | 22 |
| IQR | 6 | 12 |
| Range | 10 | 14 |
Why coaches (and gym managers) care about IQR, not just the average
Reporting stats separately instead of comparing them
Writing 'Class A mean is 23. Class B mean is 28.' as two disconnected sentences with no comparative word.
Consequence: Loses the comparison mark specifically, even though both individual values are entirely correct.
Fix: Join the two facts with a comparative word in ONE sentence: 'Class B's mean of 28 is 5 marks higher than Class A's mean of 23.'
'Compare' answers need a linking word
Key point
Key takeaways
- Modal class as a percentage (frequency ÷ total × 100) gives a far more informative interpretation than naming the class alone.
- 'Compare' demands a direct relational statement joining both datasets, covering centre AND spread wherever possible.
- Identical means or medians can hide wildly different spreads — always check IQR/range before calling two groups 'equivalent'.
Quick Reference
Statistics – Averages and Analysis: everything on one page
Formulas
- class midpoint
- always use boundaries
- grouped-data mean (an estimate)
- reverse the mean to find a missing value
- spread of the middle 50%
- crude spread, sensitive to outliers
- raw-data mean (exact)
Key facts
- Grouped mean is always...
- an estimate, never the exact mean
- Even-n median needs...
- the average of the two middle values
- IQR ignores...
- the top and bottom 25% of the data (the extremes)
Mnemonics
- Mode = Most common. Median = Middle. Mean = Maths.Three M's, three different questions about 'typical' — stops you defaulting to mean automatically.
Don't confuse
- Class limit vs Class boundary: Limit = printed table number; boundary = true dividing line, midpoint between adjacent limits (usually ends in .5).
- Mean vs Median: Mean uses every value and is pulled by outliers; median is the middle value and resists them.
- Range vs Interquartile Range: Range = max − min (whole spread, outlier-sensitive); IQR = Q3 − Q1 (middle 50%, outlier-resistant).
Command terms
- State/Identify:
- Give the single value only — no working needed, no interval where a number was asked for.
- Calculate:
- Show the method — a method mark survives even a wrong final arithmetic slip.
- Describe:
- Write a full sentence about a trend or pattern, not one word.
- Explain:
- Give the reasoning/'why', not just the fact.
- Compare:
- Use a linking word (higher/lower/more consistent) joining BOTH datasets in one statement.
Exam checklist
- Ordered the data before finding median or quartiles?
- Used boundaries (not limits) for class width?
- Divided by the TOTAL frequency/count, not the number of intervals?
- Included the unknown value in n when reverse-engineering a missing data point?
- Compared BOTH centre and spread, with a linking word, in any 'compare' answer?
- Checked whether 'no mode' is actually the correct answer before forcing one?