SAT Statistics & Data Interpretation: Tables, Means & Scatterplots

Data interpretation and basic statistics appear about 4–6 times in Digital SAT Math — and similar graph-reading skills show up in Reading & Writing. Formats: tables, bar charts, histograms, scatterplots, box plots, and short data stories with no figure at all.

Statistics content is conceptual more than computational. You will use mean, median, mode, and range. You will interpret standard deviation (larger = more spread) but not grind a full SD formula. Scatterplots often ask about association, a line of best fit, or what a slope/intercept means in context.

The points are won by careful reading: correct row, correct axis scale, correct question target — not by advanced stats theory.

Center & spread: mean, median, mode, range; effect of an outlier; which measure is pulled by skew.

Two-way tables: joint counts, row/column totals, conditional fractions (“of those who…”).

Displays: bar/histogram comparisons, box-plot five-number summaries, scatterplot trends.

Line of best fit: slope and intercept in context; predicted value; residual idea (actual − predicted) when mentioned; limits of extrapolation.

Study-adjacent wording sometimes appears near sampling items — pair with Probability & Sampling when the question is about valid conclusions rather than a calculation.

Label the format before computing. “What is the median?” is not the same skill as “which statement about the line of best fit is true?”

Data set: 4, 6, 7, 9, 24.

Mean = (4+6+7+9+24)/5 = 50/5 = 10.

Median (middle when ordered): 7. Mode: none (or all unique). Range = 24 − 4 = 20.

The 24 pulls the mean up above the median — a classic right-skew / outlier pattern. If a question asks which measure best represents a “typical” value for a skewed set, median is often preferred; if it asks for the arithmetic average, compute the mean.

Effect of change: removing 24 leaves 4, 6, 7, 9 — mean becomes 6.5, median becomes 6.5. Watch whether the question adds a value, removes one, or doubles every value (mean and median scale; range scales; “doubling” can be a trap if only some values change).

Trap: computing mean when asked for median (or the reverse) because the numbers look easy.

Suppose a survey of 200 students. STEM: 40 prefer online, 60 prefer in-person (100 total). Non-STEM: 70 online, 30 in-person (100 total). Column totals: 110 online, 90 in-person.

Q1: What fraction of all students prefer online? 110/200 = 11/20. Denominator = grand total.

Q2: What fraction of STEM students prefer online? 40/100 = 2/5. Denominator = STEM row only — not 200.

Q3: Of students who prefer in-person, what fraction are Non-STEM? 30/90 = 1/3. Denominator = in-person column.

Rule: the phrase after “of” (or “among”) names the denominator group. Underline that group before any division.

Trap: using 40/200 for Q2 because 40 is “STEM and online.” That answers a different question (joint, not conditional).

A scatterplot of study hours (x) vs. practice-test score (y) shows a positive linear trend. A line of best fit is y = 12x + 880.

Slope 12: each additional study hour is associated with about 12 more points on the model — not a proven causal law, but the modeled rate of change.

Intercept 880: the model’s predicted score when x = 0 hours. It may or may not be realistic; still answer what the intercept means if asked.

Prediction: at x = 5, y = 12(5) + 880 = 940.

If a point is (5, 910), residual ≈ actual − predicted = 910 − 940 = −30 (below the line).

Extrapolation trap: using x = 40 hours when all data sit between 0 and 10 — the SAT often asks which prediction is least reliable because it is far outside the data range.

Association ≠ causation wording: “associated with higher scores” is safer than “causes higher scores” unless the design supports cause (usually it does not).

Standard deviation (conceptual): Set A = {10, 10, 10, 10} has SD near 0. Set B = {0, 5, 15, 20} has a larger SD. Same mean can hide very different spreads — the SAT asks which set is more variable, not for a long calculation.

Box plot reminder: min, Q1, median, Q3, max. IQR = Q3 − Q1. A long whisker signals a more extreme tail, not necessarily a larger median.

Bar chart / axis traps: if the vertical axis starts at 80 instead of 0, small differences look huge. Read tick marks and units before comparing bar heights.

Histogram: height is frequency (or relative frequency) in a bin. Do not treat the tallest bar as a single data value — it is a count of values in that interval.

Speed habit: title → axis labels → units → question target → only then the numbers you need.

Read the question stem before deep-diving the figure. Know whether you need a single cell, a row total, a slope meaning, or a comparison.

For tables, mark the conditioning group (denominator) in one underline stroke.

For center measures, sort before median; do not assume mean ≈ median on skewed sets.

For fit lines, write one English sentence for slope and one for intercept with units before looking at choices.

Estimate: if a bar is roughly halfway between 40 and 60, do not invent 41.3 unless the scale supports it.

When two choices differ only by denominator (row vs total), re-read “of which group?”

Grand-total denominator on conditional questions — the #1 table miss.

Mean/median swap under time pressure.

Treating SD as a measure of center or as “average distance” you must compute with a formula.

Misreading axis scales or units (thousands, percents, years).

Extrapolating far past the data and treating the model as exact truth.

Causal language on observational scatterplots.

Reading the wrong series on a multi-bar chart (Group A vs Group B).

Day 1: Mean/median/mode/range and outlier effects only.

Day 2: Two-way tables — joint vs conditional fractions only.

Day 3: Bar charts and histograms with deliberate axis-scale traps.

Day 4: Scatterplots + slope/intercept interpretation (no hard algebra).

Day 5: Box plots and SD comparison items.

Day 6: Mixed 12–15; tag every miss as “wrong group,” “wrong measure,” or “misread display.”

Day 7: Timed mini-set (8 items in ~10 minutes), then untimed figure re-read of every miss.

Ready bar: you can name the denominator group in one second and explain slope of a fit line in unit-aware English.

Data Analysis topic sessions mix tables, charts, and center/spread items so you practice format recognition under mixed order — closer to a real module than a single worksheet type.

After a miss, ask the AI tutor (Pro) “what was the denominator group?” or “explain the slope in context only.” Narrow prompts fix the exact leak.

Pair with Ratios & Proportions when fractions live inside tables, and with Probability & Sampling when the item shifts from calculation to study design and valid conclusions.