GRE Quant lesson · Distributions & quartiles
GRE distributions: quartiles, boxplots, percentiles and the normal distribution
How to handle GRE distribution questions: quartiles and the interquartile range, reading boxplots, percentile rank from a table, the 68-95-99.7 rule for normal distributions, and comparing z-scores.
The core idea
Sort the data and split it in halves: the quartiles are the medians of the halves, and the IQR (Q3 − Q1) is the spread of the middle half. A boxplot draws those five numbers but never the mean. For a normal distribution, measure every cutoff in standard deviations from the mean and use 68-95-99.7.
Core ideas
Words used here
- Quartiles: the three values that cut a sorted list into four equal parts. Q1 (first quartile), the median (Q2) and Q3 (third quartile).
- Interquartile range (IQR): Q3 − Q1, the spread of the middle half of the data.
- Boxplot (box-and-whisker plot): a drawing of five numbers on a scale: least value, Q1, median, Q3, greatest value.
- Percentile: the Pth percentile is a value with about P percent of the data at or below it. The percentile rank of a value is the percent of the data at or below it.
- Normal distribution: the symmetric bell-shaped distribution, centred at its mean.
- Standard deviation (SD): the typical distance of the values from the mean.
- z-score: (value − mean) ÷ SD, the number of SDs a value is above (+) or below (−) the mean.
- Quartiles by halves. Sort the list and split it into a lower and an upper half (with an odd count, leave the median out of both). Q1 is the median of the lower half, Q3 the median of the upper half.
- IQR = Q3 − Q1; range = greatest − least.
- Percentile rank = (number of values at or below) ÷ (total) × 100.
- 68-95-99.7. In a normal distribution about 68% of values lie within 1 SD of the mean, 95% within 2 SDs and 99.7% within 3, split evenly on the two sides.
On the GRE.
- Calculator: the on-screen calculator helps with a percentile like 26 ÷ 40 × 100, but most distribution questions need sorting, splitting and the 68-95-99.7 bands, not arithmetic.
- Quantitative Comparison: boxplots and normal distributions are QC favourites. Ask first whether the display or the information fixes the quantity at all: a boxplot fixes medians and quartiles, never the mean.
- Numeric Entry: type percents as numbers in the box (65 for 65%) and keep halves exactly (13.5, not 14) unless the question says to round.
Try it. Find Q1, the median, Q3 and the IQR of 3, 5, 7, 8, 10, 12, 15, 20.
Show the answer
- 8 values, already sorted. Lower half 3, 5, 7, 8; upper half 10, 12, 15, 20.
- Q1 = (5 + 7) ÷ 2 = 6. Median = (8 + 10) ÷ 2 = 9. Q3 = (12 + 15) ÷ 2 = 13.5.
- IQR = 13.5 − 6 = 7.5. (The range is 20 − 3 = 17.)
Quartiles and the interquartile range
Sort the data, cut it into two halves, and take the median of each half. Those medians are Q1 and Q3, and the IQR is Q3 − Q1. About half the data lies between Q1 and Q3.
Why it works. The median splits the data into two equal halves; the medians of those halves split each half again, so the three values cut the data into quarters. The IQR measures the spread of the middle two quarters, so one extreme value can't stretch it the way it stretches the range.
How to spot it: "interquartile range", "first quartile", "third quartile", "the middle half of the data".
Example. Find the IQR of 4, 6, 7, 9, 11, 12, 14, 15, 18, 21, 25, 30.
- 12 values, sorted. Lower half: 4, 6, 7, 9, 11, 12. Upper half: 14, 15, 18, 21, 25, 30.
- Q1 = (7 + 9) ÷ 2 = 8 and Q3 = (18 + 21) ÷ 2 = 19.5.
- IQR = 19.5 − 8 = 11.5.
Trap: answering the range, 30 − 4 = 26, or Q3 alone, 19.5.
Edge cases.
- Odd count: leave the median out of both halves. For 2, 4, 5, 8, 9, 11, 15 the median is 8, Q1 = 4 (middle of 2, 4, 5) and Q3 = 11, so the IQR is 7.
- Unsorted lists: sort first; the middle of the printed list means nothing.
- Repeated values stay in the list: a quartile can equal the median.
GRE-level example. For the list 10, 12, 12, 15, 17, 20, 26, 40, which is greater: the range, or 3 times the IQR?
- Q1 = (12 + 12) ÷ 2 = 12 and Q3 = (20 + 26) ÷ 2 = 23, so the IQR is 11 and 3 × IQR = 33.
- The range is 40 − 10 = 30.
- 3 times the IQR is greater. The one far value, 40, stretches the range but barely moves the IQR.
Check yourself. What is the IQR of 1, 3, 3, 6, 7, 8, 9, 13?
Check your answer
Lower half 1, 3, 3, 6: Q1 = 3. Upper half 7, 8, 9, 13: Q3 = 8.5. IQR = 8.5 − 3 = 5.5.
Reading a boxplot
A boxplot shows five numbers against a scale. The whiskers end at the least and greatest values, the box runs from Q1 to Q3, and the line inside the box is the median. GRE boxplots are drawn to scale, so read each value from the number line.
Why it works. The five numbers cut the data into four parts with about a quarter of the values in each: least to Q1, Q1 to the median, the median to Q3, Q3 to greatest. A long whisker or a wide half-box means that quarter is spread out, not that it holds more values.
How to spot it: a box with whiskers over a number line; quantities such as "the IQR of group P" or "the mean of group Q".
Example. A boxplot shows least value 20, Q1 35, median 45, Q3 50 and greatest value 80.
- The IQR is the box's width: 50 − 35 = 15. The range is 80 − 20 = 60.
- About half the values lie between 35 and 50, and about a quarter lie between 50 and 80, even though that whisker is long.
- The mean isn't shown. It lies between 20 and 80, but where depends on data the plot doesn't show.
Trap: reading the median line as the mean.
Edge cases.
- Comparing two boxplots: a box farther right has bigger values; a wider box has a bigger IQR. These are different questions.
- The mean against a number outside the whiskers can be decided: every value is at least the least value, so the mean is too.
- A long whisker means spread-out values in that quarter, not more values.
GRE-level example. Using the boxplot above (20, 35, 45, 50, 80), compare Quantity A, the mean of the data, with Quantity B, 45.
- The plot fixes the five numbers, not the mean.
- A data set with many values near 20 and 35 can have a mean below 45; one with many values near 50 and 80 can have a mean above it, and both fit the same plot.
- The answer is (D). Against 15 instead, the answer would be (A): every value is at least 20.
Check yourself. Two boxplots: P has Q1 = 30 and Q3 = 42; Q has Q1 = 50 and Q3 = 58. Which has the greater IQR?
Check your answer
P: 42 − 30 = 12, against 58 − 50 = 8 for Q. Q's box sits farther right, but it is narrower.
Percentiles and percentile rank
The percentile rank of a value is the percent of the data at or below it. From a frequency table, add the counts down to and including that value's row, divide by the total, and multiply by 100.
Why it works. A percentile describes position, not score: it says what share of the group did no better. That is why a score of 8 out of 10 (80% correct) can be at the 65th percentile.
How to spot it: "percentile", "what percent of the students scored … or less", a frequency table with a total.
Example. Quiz scores of 40 students:
| Score | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|
| Students | 2 | 4 | 8 | 12 | 9 | 5 |
What percent scored 8 or less?
- Counts at or below 8: 2 + 4 + 8 + 12 = 26.
- 26 ÷ 40 × 100 = 65%. So a score of 8 has percentile rank 65.
Trap: "less than 8" leaves out the 12 students who scored 8: 14 ÷ 40 = 35%. Read the boundary.
Edge cases.
- Percentile ≠ percent correct. 8 out of 10 is 80% correct; here it is the 65th percentile.
- "At least" and "more than" count from the top: 8 or more is 12 + 9 + 5 = 26 students, 65%; more than 8 is 14 students, 35%.
- The Pth percentile from a table is the first value whose running total reaches P% of the count.
GRE-level example. In the table above, which score is at the 90th percentile?
- Running totals: 2, 6, 14, 26, 35, 40, which are 5%, 15%, 35%, 65%, 87.5% and 100%.
- 87.5% of students scored 9 or less, short of 90%; 100% scored 10 or less.
- The 90th percentile is 10.
Check yourself. What is the percentile rank of a score of 7 in the table above?
Check your answer
2 + 4 + 8 = 14 students scored 7 or less: 14 ÷ 40 × 100 = 35.
The normal distribution: 68-95-99.7
For a normal distribution with mean m and standard deviation s, about 68% of the values lie between m − s and m + s, 95% between m − 2s and m + 2s, and 99.7% between m − 3s and m + 3s. Because the curve is symmetric, each percent splits evenly on the two sides of the mean.
Why it works. Every normal distribution is the same bell curve, stretched by its SD and shifted to its mean. So the share of values within k SDs of the mean is the same for all of them; 68, 95 and 99.7 are those shares for k = 1, 2, 3.
How to spot it: "normally distributed", "approximately normal", a mean and a standard deviation, "approximately what percent".
Example. Exam scores are normally distributed with mean 70 and SD 5. About what percent are between 65 and 80?
- 65 is 1 SD below the mean; 80 is 2 SDs above.
- Bands: 65 to 70 is 34%, 70 to 75 is 34%, 75 to 80 is (95 − 68) ÷ 2 = 13.5%.
- Total: 34 + 34 + 13.5 = 81.5%.
Trap: answering 68% for "between the mean and 1 SD above it". That is one side only: 34%.
Edge cases.
- Above m + s: 50% − 34% = 16%. Above m + 2s: 2.5%.
- Below the mean: exactly 50%, the line of symmetry.
- Probability language: "the probability that a randomly chosen value is between …" is the same share, as a decimal (0.815).
GRE-level example. The lifetimes of 1,200 bulbs are normally distributed with mean 1,000 hours and SD 80 hours. About how many last between 920 and 1,160 hours?
- 920 is 1 SD below the mean; 1,160 is 2 SDs above.
- Share: 34 + 34 + 13.5 = 81.5%.
- 0.815 × 1,200 = 978 bulbs.
Check yourself. For a normal distribution with mean 40 and SD 6, about what percent of values are less than 28?
Check your answer
28 is 2 SDs below the mean (40 − 12), so about 2.5%.
Comparing normal distributions with z-scores
To compare tails of two different normal distributions, convert each cutoff to a z-score: (cutoff − mean) ÷ SD. Tails beyond the same z-score are the same size, and a cutoff more SDs from the mean leaves a smaller tail.
Why it works. Subtracting the mean and dividing by the SD turns every normal distribution into the same standard one. A cutoff's z-score is its position on that shared curve, so tails can be compared without any table.
How to spot it: two normal variables with different means and SDs in one QC; "the probability that a value from X is greater than …" against the same for Y.
Example. X is normal with mean 50 and SD 4; Y is normal with mean 80 and SD 10. Quantity A: P(X > 56). Quantity B: P(Y > 92).
- For X: (56 − 50) ÷ 4 = 1.5 SDs above the mean.
- For Y: (92 − 80) ÷ 10 = 1.2 SDs above the mean.
- 1.5 SDs out leaves the smaller tail, so Quantity B is greater (B).
Symmetry turns "less than" around. P(Y < 68) is the tail more than (80 − 68) ÷ 10 = 1.2 SDs below the mean, the same size as P(Y > 92).
Trap: comparing the raw distances 6 and 12. The bigger distance is fewer SDs here.
Edge cases.
- Same z, same tail: P(X > 54) = P(Y > 90), since both are 1 SD above.
- A tail toward the mean (P(X > 46) when the mean is 50) is more than 50%.
- An unknown distance k: k above the mean is k ÷ 4 SDs for X and k ÷ 10 for Y. For k > 0, X's tail is smaller; if k could be negative, the order flips.
GRE-level example. With X and Y as above and k a nonzero number, compare Quantity A, P(X > 50 + k), with Quantity B, P(Y > 80 + k).
- The cutoffs are k/4 and k/10 SDs above the means.
- If k = 4: 1 SD and 0.4 SDs, so B is greater. If k = −4: −1 and −0.4 SDs, so A is greater (its cutoff is further below the mean).
- The answer is (D). If the question said k > 0, it would be (B).
Check yourself. X: mean 30, SD 5. Y: mean 60, SD 15. Which is greater, P(X < 20) or P(Y < 30)?
Check your answer
They are equal: 20 is (30 − 20) ÷ 5 = 2 SDs below X's mean, and 30 is (60 − 30) ÷ 15 = 2 SDs below Y's mean.
How hard questions disguise it
For harder questions (L4–L5)
L4 questions test what a display can't tell you: the mean from a boxplot, or a tail beyond an unknown distance k. Before computing, ask whether the information fixes the quantity at all, and build two allowed cases when it doesn't. L5 questions combine the normal distribution with probability ("the probability that a randomly chosen value is not between …" is 1 minus the share inside) or with standard deviation (a value's distance from the mean in SDs). Convert everything to SDs from the mean first, then add the 68-95-99.7 bands.
Practice
Practice these in timed sets on the distributions practice page.
Common traps
Mixing up the IQR, the range and the quartiles
The IQR is Q3 − Q1, the width of the box; the range is greatest − least, whisker to whisker. A quartile on its own, or the distance from the median to one quartile, is not the IQR.
Taking a quartile from the wrong position
Sort first. Q1 is the median of the lower half and Q3 the median of the upper half; with 12 values each half has 6, so each quartile is the average of the 3rd and 4th values of its half.
Reading the mean from a boxplot
The line inside the box is the median. A boxplot never shows the mean: different data with the same boxplot have different means, so a comparison with the mean is usually (D) unless the value is outside the whiskers.
Judging spread by where a box sits
A box farther right holds bigger values, not a bigger spread. Compare widths for the IQR and whisker spans for the range.
Using the wrong part of the 68-95-99.7 rule
68% is within 1 SD on both sides together, so 34% on each side; 95% within 2 SDs leaves 2.5% in each tail. Turn the cutoffs into SDs from the mean, sketch the six bands, and add the ones you need.
Comparing raw distances instead of z-scores
6 points above a mean is 1.5 SDs if the SD is 4 but only 0.6 SDs if it is 10. Divide each distance by its own SD before comparing tails.