GMAT Data Insights lesson · Correlation
GMAT Data Insights: scatter plots and correlation
Reading scatter plots on GMAT Data Insights: direction of a relationship, counting dots in a region, conditional fractions, and tendencies vs rules.
The core idea
Judge a relationship from the whole cloud of dots. Count regions with two imaginary lines. A conditional fraction's denominator is the group the condition picks out. A tendency is not a rule.
Core ideas
Each dot is one entity with two values. Questions ask about the cloud (direction), about a region (counts), or about one dot (reading).
A scatter plot is a table with two columns drawn as dots. Every dot has an across value (x, the horizontal axis) and an up value (y, the vertical axis):
| Shop | Staff (x, across) | Sales $K (y, up) |
|---|---|---|
| Ash | 3 | 40 |
| Elm | 5 | 55 |
| Oak | 7 | 50 |
- The rightmost dot has the greatest x: Oak (7 staff).
- The highest dot has the greatest y: Elm ($55K).
- They are different dots. Name the axis before you answer.
The question types and what to do:
| Question | Look at | Do |
|---|---|---|
| Direction | the whole cloud | does it rise or fall to the right? |
| Region count | two cut-off lines | count one corner |
| Conditional fraction | the group named after "of" | count it first, it's the denominator |
| Rule in a table | every pair of rows | sort by one column, check the other |
- Name what each axis measures and its units.
- For a region, picture a vertical line at the x cut-off and a horizontal line at the y cut-off.
- Count the dots in the corner the statement names. Dots exactly on a line need the threshold words ("above" excludes them).
- For "of the …", count that group first; it is the whole.
Direction
Why it works. A positive relationship says that larger x values tend to come with larger y values. It is a claim about the whole group, not about each pair of dots. Comparing the left part of the cloud with the right part lets the small dips cancel out, so a few dots can't decide it.
How to spot it: "The chart shows ___ relationship between floor area and sales."
Dots rising to the right mean higher x goes with higher y: a positive relationship. A few dots lining up don't make a trend if the rest don't follow.
Example. Ten stores. Each point is one store: floor area (hundred m²) across, monthly sales ($ thousand) up. Here the stores happen to be evenly spaced in floor area, so the points are read from left to right.
- Compare the left end with the right end: the smallest stores sell about $50–62K, the largest $96–102K.
- Check the middle: the cloud keeps climbing even though 4 of the 9 steps dip.
- The relationship is positive.
Compare a cloud with no clear direction:
Trap: judging from two or three dots. In the first chart, the stores at 14 and 16 fall from 82 to 76, but the whole cloud still rises. In the second chart, the 3rd and 5th points (85 and 90) line up upward, yet the rest don't follow, so there is no clear relationship.
Try it. Six students: hours of study across, errors on a test up. What is the direction of the relationship?
| Hours (x) | 2 | 4 | 6 | 8 | 10 | 12 |
|---|---|---|---|---|---|---|
| Errors (y) | 18 | 15 | 16 | 11 | 9 | 7 |
Show the answer
- Compare the ends: the two students who studied least made 15–18 errors, the two who studied most made 7–9.
- Check the middle: 4 of the 5 steps fall. The one rise (15 to 16) doesn't change the overall picture.
- More hours go with fewer errors: a negative relationship.
Edge cases.
- Direction is not steepness. A cloud that rises slowly is still positive.
- An arch has no single direction. If the dots rise and then fall, neither "positive" nor "negative" describes the whole range.
- A relationship is not a cause. The chart shows that two values move together, not that one makes the other change.
GMAT-level example. Six apartment buildings: distance from the train station across, average monthly rent up. Which is supported: "Rent tends to fall with distance", "Every building farther from the station has a lower rent", or both?
| Building | Distance (km, x) | Rent ($, y) |
|---|---|---|
| Arden | 0.4 | 2,150 |
| Brook | 0.9 | 1,980 |
| Crest | 1.3 | 2,040 |
| Dover | 2.0 | 1,760 |
| Elms | 2.6 | 1,690 |
| Forde | 3.5 | 1,520 |
- The rows are sorted by distance, so read the rents from top to bottom: 2,150, 1,980, 2,040, 1,760, 1,690, 1,520.
- Compare halves. The 3 nearest buildings average 6,170 ÷ 3 ≈ $2,057. The 3 farthest average 4,970 ÷ 3 ≈ $1,657. The far half is $400 cheaper, so rent tends to fall with distance: a negative relationship.
- Check every step for the "every" claim: 4 of the 5 steps fall, but Crest (1.3 km) costs $60 more than Brook (0.9 km). One exception breaks it.
- Only the first statement is supported. ✓ Check: the ends agree with the halves (2,150 at 0.4 km vs 1,520 at 3.5 km).
Check yourself. Five dots, read from left to right, have y values 10, 14, 13, 18 and 21. What is the direction?
Check your answer
3 of the 4 steps rise, and the right end (21) is far above the left end (10): positive. The dip from 14 to 13 is noise.
Counting a region
Why it works. Each condition is a line that splits the dots into two sides. Two lines make four corners, and every dot sits in exactly one of them. An "and" question names one corner, so you count each dot once or not at all.
How to spot it: "How many stores have floor area above 16 and monthly sales above $80K?"
"How many stores have floor area above 16 and monthly sales above $80K": picture a vertical line at 16 and a horizontal line at 80, and count the dots right of the first and above the second.
Example. The same ten stores. Count the stores with floor area above 16 and sales above 80.
- Right of 16 (above 16, so the store at exactly 16 is out): areas 18, 20, 22, 24, 26. That's 5 stores.
- Of those, sales above 80: 88, 84, 96, 102. The store at 18 has only 74.
- Count: 4.
| Area > 16 | Area ≤ 16 | |
|---|---|---|
| Sales > 80 | 4 | 1 |
| Sales ≤ 80 | 1 | 4 |
Trap: using one condition only gives 5 (all stores right of 16, or all above 80). Counting either condition (OR) gives 6. The store at 14 with sales 82 is above the line but on the wrong side of 16.
Try it. Six students: weekly practice hours and test score. How many practiced more than 4 hours and scored below 70?
| Hours | 2 | 4 | 5 | 6 | 8 | 9 |
|---|---|---|---|---|---|---|
| Score | 55 | 72 | 68 | 74 | 65 | 88 |
Show the answer
- More than 4 hours: 5, 6, 8 and 9 hours. That's 4 students. The student at exactly 4 is out.
- Their scores: 68, 74, 65, 88. Below 70: 68 and 65.
- Count: 2.
Edge cases.
- "At least" and "at most" include the dots on the line; "above", "more than" and "below" leave them out.
- An "or" count is A + B − both, because the corner in both groups would otherwise be counted twice.
- "Neither" is the total minus the "or" count: the one corner outside both lines.
GMAT-level example. Six delivery routes: distance (km) and average delivery time (minutes). How many routes satisfy exactly one of these: distance above 30 km, time above 60 minutes?
| Route | R1 | R2 | R3 | R4 | R5 | R6 |
|---|---|---|---|---|---|---|
| Distance (km) | 12 | 25 | 30 | 34 | 40 | 52 |
| Time (min) | 35 | 42 | 66 | 58 | 75 | 80 |
- Distance above 30: R4, R5, R6 (3 routes). R3 is exactly 30, so it's out.
- Time above 60: R3, R5, R6 (3 routes).
- Both: R5 and R6. Only distance: R4 (58 minutes). Only time: R3 (exactly 30 km).
- Exactly one = only distance + only time = 1 + 1 = 2.
| Distance > 30 | Distance ≤ 30 | |
|---|---|---|
| Time > 60 | 2 | 1 |
| Time ≤ 60 | 1 | 2 |
✓ Check: 2 + 1 + 1 + 2 = 6 routes. Treating R3 as "above 30" moves it into "both" and gives 1, a trap answer.
Check yourself. In the Try it table, how many students practiced more than 4 hours or scored above 70?
Check your answer
More than 4 hours: 4. Above 70: 3 (72, 74, 88). Both: 2 (6 and 9 hours). Or = 4 + 3 − 2 = 5. Only the student with 2 hours and 55 points is out.
Conditional fractions
Why it works. "Of the plots with more than 140" tells you to set every other plot aside, so the group it names becomes the whole. A fraction is part over whole: the "of" condition sets the denominator, and the second condition picks the part inside it.
How to spot it: "Of the plots that received more than 140 kg/ha of fertilizer, the fraction with yield above 80 is ___."
Example. Eleven test plots: fertilizer (kg per hectare) and yield (hundred kg per hectare).
| Plot | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Fertilizer | 100 | 110 | 120 | 130 | 135 | 145 | 150 | 160 | 170 | 180 | 190 |
| Yield | 60 | 72 | 81 | 85 | 78 | 82 | 90 | 76 | 88 | 95 | 84 |
- "Of the plots with more than 140" makes those plots the whole. They are plots 6 to 11: 6 plots.
- Among those 6, yield above 80: 82, 90, 88, 95, 84. That's 5 (plot 8 has 76).
- The fraction is 5/6.
Trap: 5/11 divides by all the plots. Reversing the condition, "of the plots with yield above 80, the fraction with more than 140 kg", is a different question: 7 plots yield above 80, so it's 5/7. And 5 : 1 (above vs not above) is part-to-part.
Try it. Use the eleven plots above. Of the plots with yield above 85, what fraction received more than 150 kg/ha?
Show the answer
- The whole: plots with yield above 85. They are plot 7 (90), plot 9 (88) and plot 10 (95). That's 3 plots.
- The part: of those, more than 150 kg/ha. Plot 7 has exactly 150, so it's out. Plots 9 and 10 are in: 2 plots.
- The fraction is 2/3. (Dividing by all 11 plots gives 2/11, which is wrong.)
Edge cases.
- Reversing the condition changes the whole. "Of A, the fraction that are B" and "of B, the fraction that are A" share a part but not a denominator.
- "What fraction of all plots are A and B" is a different question: its denominator is every dot.
- An empty group has no fraction. If no dot meets the "of" condition, the fraction doesn't exist; it isn't 0.
GMAT-level example. Six salespeople: years of experience and annual sales. Of the salespeople with sales above $400K, what fraction have more than 5 years of experience? Is it greater or less than the fraction of the more-than-5-years group with sales above $400K?
| Salesperson | Ana | Ben | Cai | Dev | Eli | Fay |
|---|---|---|---|---|---|---|
| Experience (years) | 2 | 4 | 5 | 7 | 9 | 12 |
| Sales ($K) | 310 | 420 | 440 | 450 | 380 | 490 |
- Whole for the first fraction: sales above 400. Ben, Cai, Dev and Fay: 4 people.
- Part: of those 4, more than 5 years. Dev and Fay: 2. The fraction is 2/4 = 1/2.
- Whole for the second fraction: more than 5 years. Dev, Eli and Fay: 3 people. Of them, sales above 400: Dev and Fay, 2. The fraction is 2/3.
- 1/2 < 2/3, so the first fraction is less. Same part (Dev and Fay), different wholes (4 vs 3). ✓ Check: dividing by all 6 gives 2/6 = 1/3, a trap answer.
Check yourself. In the salespeople table, of those with 5 years or less, what fraction have sales above $400K?
Check your answer
The whole is Ana, Ben and Cai: 3 people. Ben and Cai sell above 400, so the fraction is 2/3.
Rules in a table
Why it works. "Every" is a claim about each pair of rows. Once the rows are sorted by one column, the other column must go up at every neighbouring step, and steps that all go up chain together across any pair. So checking neighbours is enough, and one step down is a counterexample.
How to spot it: "Every warehouse with more employees has more floor area" (true or false).
"Every warehouse with more employees has more floor area" is true only if, sorted by employees, floor area rises at every step.
Example. Six warehouses, already sorted by employees.
| Warehouse | Employees | Floor area (thousand m²) |
|---|---|---|
| Avon | 40 | 12 |
| Bray | 55 | 15 |
| Cole | 60 | 18 |
| Dale | 72 | 17 |
| Esk | 85 | 22 |
| Fyne | 90 | 25 |
- Sort by employees (done here).
- Check floor area at every step: 12 → 15 → 18 → 17 → 22 → 25.
- Dale has more employees than Cole (72 vs 60) but less floor area (17 vs 18). The rule is false.
Trap: the trend is clearly positive, and the warehouse with the most employees (Fyne) does have the most floor area. Neither makes "every" true. One exception is enough to break a rule.
Try it. True or false: "Every café with more seats has more daily customers"?
| Café | Bean | Cup | Dock | Fig | Gale |
|---|---|---|---|---|---|
| Seats | 30 | 18 | 42 | 24 | 36 |
| Customers a day | 210 | 150 | 260 | 190 | 245 |
Show the answer
- Sort by seats: Cup 18, Fig 24, Bean 30, Gale 36, Dock 42.
- Read customers in that order: 150, 190, 210, 245, 260.
- Customers rise at every step, so the rule is true.
Edge cases.
- "More" is strict. Two rows with different x and the same y break "has more".
- A rule can hold on part of a table. In the warehouse table it holds from Avon to Cole (12, 15, 18) and from Dale to Fyne (17, 22, 25), but not across the break.
- "The most A also has the most B" checks only the top row. It is a much weaker claim than "every".
GMAT-level example. Six laptops. True or false: (1) every laptop that costs more has a longer battery life; (2) every laptop that costs more weighs less?
| Laptop | P | Q | R | S | T | U |
|---|---|---|---|---|---|---|
| Price ($) | 650 | 900 | 720 | 1,100 | 820 | 1,300 |
| Battery (hours) | 7 | 10 | 8 | 12 | 9 | 11 |
| Weight (kg) | 1.9 | 1.5 | 1.8 | 1.3 | 1.6 | 1.2 |
- Sort by price: P 650, R 720, T 820, Q 900, S 1,100, U 1,300.
- Battery in that order: 7, 8, 9, 10, 12, 11. U costs more than S but lasts less (11 vs 12). Statement 1 is false.
- Weight in that order: 1.9, 1.8, 1.6, 1.5, 1.3, 1.2. It falls at every step. Statement 2 is true.
- ✓ Check: both columns follow the price trend, but only weight follows it without an exception.
Check yourself. In the laptop table, is it true that every laptop with a longer battery life weighs less?
Check your answer
Sort by battery: P 7, R 8, T 9, Q 10, U 11, S 12. Weights: 1.9, 1.8, 1.6, 1.5, 1.2, 1.3. S lasts longer than U but weighs more (1.3 vs 1.2): false.
How hard questions disguise it
For harder questions (L4–L5)
At L4 and L5, a scatter plot rarely asks only "positive or negative". It hides a condition count behind a boundary value that sits exactly on a cut-off line, or asks for a share of a subgroup whose denominator is set by "of the …". A statement may sound like a trend but use "every" or "each", so it is really a rule you test row by row after you sort the table. Others mix the cloud with a single dot: "the store with the greatest sales also has the greatest floor area" is about two read values, not the trend. Some ask whether a dot lies above or below the general trend. Name the axis, the cut-offs and the whole before you count.
Practice
Practice scatter plot and correlation questions in timed sets on the correlation practice page.
Common traps
Reading the direction wrong
Rising to the right is positive; falling is negative; no consistent direction is no clear relationship. Look at all the dots, not two or three.
Counting the wrong region
Draw both cut-off lines in your head and count only the corner that satisfies both conditions.
The wrong denominator
'Of the stores with floor area above 16' makes those stores the whole: count them first.
A trend treated as a rule
'Every row with more A has more B' fails on a single exception. Sort by A and check B at every step.
Reading the wrong dot or axis
The rightmost dot has the greatest horizontal value; the highest dot has the greatest vertical value.