7 topics. Each one opens with its key idea, then the detail you need to use it.
1. Ratios, unit rates, and proportions
Key idea: Find the unit rate or set up a proportion, keep the units attached, and never round the multiplier.
A ratio compares two quantities, like 3 cups of flour to 2 cups of milk. A rate is a ratio with different units, like 240 miles per 4 hours, and the unit rate scales it down to one of the second unit: 240 ÷ 4 = 60 miles per hour. Once you have a unit rate, any other amount is one multiplication away, so 7 hours at that speed gives 420 miles.
A proportion sets two ratios equal. If 4 notebooks cost $10, then 4/10 = 14/x for 14 notebooks, and cross-multiplying gives 4x = 140, so x = 35 dollars. The unit-rate route gives the same answer, since $2.50 per notebook times 14 is $35, so use whichever route fits the numbers.
Fractions show up inside rates. If a recipe needs 3/4 cup of oil per batch and you make 2.5 batches, multiply (3/4) × (5/2) = 15/8, which is 1 7/8 cups. Rewrite the decimal as a fraction first so the arithmetic stays exact. Do not round 2.5 batches to 2 or 3, because that gives the oil for a different number of batches.
Unit conversions use the same idea. Multiply by a conversion factor arranged so the unit you want to cancel sits on the bottom, so 3 ft × 12 in/ft = 36 in. For two steps, chain the factors. A price of $4.80 per pound is 480 cents per pound, and since a pound is 16 ounces, 480 ÷ 16 = 30 cents per ounce.
Two checks catch most errors. Track units, so miles per hour times hours leaves miles, and then check direction, because more batches must need more oil. Also watch part-to-part versus part-to-whole. If the ratio of boys to girls is 2:3, the boys are 2/5 of the class, not 2/3.
2. Percentages and percent change
Key idea: Multiply by the decimal, check which group is asked for, and always measure percent change against the original.
Percent means per hundred, so 35% is 0.35. To find a percent of a total, convert it to a decimal and multiply: 35% of 260 is 0.35 × 260 = 91. For a mental check, find 10% first, which is 26, so 35% is 3.5 × 26 = 91.
Read which group the question wants. If 35% of 260 people own a bike, then 91 do and 169 do not, and test writers put both numbers in the choices. They also offer the bare percent, 35, which is what you get if you forget to apply it to the total.
Percent change compares the change to the starting value: (new − old) ÷ old × 100. A price that rises from $40 to $50 went up 10 ÷ 40 = 25%, but going back down from $50 to $40 is a 10 ÷ 50 = 20% drop. The base is always the starting value, which is why the two percents differ.
Chained percents multiply. A 20% increase multiplies by 1.20 and a 20% decrease multiplies by 0.80, so raising $100 by 20% and then cutting it by 20% leaves $96, not $100. For a percent of a percent, work in steps. If 70% of 500 students play a sport and 40% of those athletes play soccer, first find 0.70 × 500 = 350, then 0.40 × 350 = 140. The second percent applies to the 350 athletes, not to all 500 students.
3. Mean, median, mode, and range
Key idea: Sort before finding a median, read frequencies as counts not values, and turn every mean into a total.
These four descriptive statistics each summarize a data set with one number. The mean is the total divided by the count, and the median is the middle value once the data is in order. The mode is the value that appears most often, and the range is the largest value minus the smallest.
For the median, sort first, then count positions. With an odd number of values n, the median is value number (n + 1)/2, so with 9 values it is the 5th. With an even count, average the two middle values, so for 3, 5, 8, 10, 12, 15 the median is (8 + 10)/2 = 9.
Frequency tables pack data tighter. Say 4 students own 0 pets, 9 own 1, 5 own 2, and 2 own 3. The mode is 1 pet, because that value has the highest frequency. It is not 9, which is only how many students gave that answer. There are 20 students, so the median is the average of the 10th and 11th values, and since values 5 through 13 are all 1s, the median is 1. The mean is (0 + 9 + 10 + 6) ÷ 20 = 1.25.
The most tested move is the mean through totals. If 20 runners average 30 minutes, their total is 600. If 4 runners who each took 40 minutes drop out, the total falls by 160 to 440, and the new mean is 440 ÷ 16 = 27.5. Divide by the new count, not the old one.
4. Outliers, spread, and standard deviation
Key idea: Outliers move the mean and range a lot and the median very little; compare standard deviations by how bunched the data is.
An outlier is a value far from the rest of the data, and it pulls on some statistics much harder than others. The mean uses the size of every value, so one huge value drags it. The range uses only the two extremes, so an outlier changes it directly, while the median depends only on position and barely moves.
Take five salaries: $40k, $42k, $45k, $47k, and $50k. The mean is $44.8k and the median is $45k. Now replace $50k with $250k. The median stays at $45k, but the mean jumps to $84.8k and the range grows from $10k to $210k. That is why the median is the better summary of a typical salary when a few values are extreme.
Standard deviation measures how far values typically sit from the mean. The PSAT rarely asks you to calculate it, but it does ask you to compare it. Data bunched near the mean has a small standard deviation, and data pushed toward the ends has a large one. The sets 10, 10, 10, 10 and 1, 5, 15, 19 both have a mean of 10, but the first has a standard deviation of 0 while the second is far more spread out.
When a value is added or removed, ask where it sits. Removing a value above the mean lowers the mean, and removing a value equal to the mean leaves it unchanged. Removing the most extreme value never increases the range. It shrinks the range unless another value ties it, and it usually lowers the standard deviation.
5. Probability, counting, and two-way tables
Key idea: Total first, then match the numerator to the question; 'given' shrinks the denominator, and independent events multiply.
Probability is favorable outcomes divided by total outcomes, when every outcome is equally likely. A jar with 5 lemon, 3 cherry, and 4 grape candies holds 12 candies, so P(cherry) = 3/12 = 1/4. Add up the total first. Wrong choices often give the probability of a different flavor, so check that your numerator matches the one the question names.
Counting matters when conditions stack. A spinner numbered 1 through 10 has 10 outcomes, and 'odd and greater than 5' means only 7 and 9, so the probability is 2/10 = 1/5. When the set is small, list the outcomes, because a list is faster than a formula and harder to get wrong.
A two-way table sorts one group by two categories. Picture 120 students sorted by grade and by whether they drive to school. Suppose 50 are juniors and 20 of those juniors drive. P(drives, given junior) = 20/50 = 2/5, because 'given' shrinks the whole to the junior row. P(junior and drives) = 20/120 = 1/6, because 'and' keeps the whole table as the denominator. The phrases 'given that', 'of those', and 'among' all signal a smaller whole.
Independent events do not affect each other, so you multiply their probabilities. Roll a die twice, and P(even both times) = (3/6) × (3/6) = 1/4. Drawing without replacement is not independent, because the first draw changes what is left. Two red cards from 4 red and 6 black gives (4/10) × (3/9) = 2/15.
6. Scatter plots, lines of best fit, and linear models
Key idea: Slope is the average rate of change, the intercept is the starting value, and every prediction comes from substituting into the line.
A scatter plot shows paired data, with one point for each person or object. The line of best fit is the straight line that runs closest to the points overall. Its slope is a rate, the average change in y for each one-unit increase in x. Its y-intercept is the predicted y when x is 0, which may or may not make sense in context.
Say a line of best fit for plant height is h = 1.5d + 4, where d is days and h is centimeters. The slope says the plants grow about 1.5 cm per day, and the intercept says a plant is predicted to be 4 cm tall on day 0. To predict, substitute: at d = 10, h = 19 cm. Stems use words like 'predicted', 'estimated', and 'on average' because the points scatter around the line rather than sitting on it.
When the data is a clean table, build the linear model yourself. If x goes 1, 3, 5, 7 and y goes 20, 26, 32, 38, then y rises 6 for every 2 units of x, so the slope is 3. Step back one unit from (1, 20) to get the intercept 17, which gives y = 3x + 17, and at x = 4 the model predicts y = 29.
A residual is the actual value minus the predicted value. A point above the line has a positive residual, which means the model underestimated it. Be cautious about predictions far beyond the plotted data, because the trend may not continue there.
7. Sampling and experimental design
Key idea: Random sampling earns generalization to the sampled population; random assignment earns cause and effect.
These questions ask what a study lets you conclude, and two features decide it. Random sampling lets you generalize from the sample to the population it was drawn from. Random assignment to treatment groups lets you conclude cause and effect.
A survey of 300 randomly chosen students at one high school supports claims about that school's students. It does not support claims about every teenager in the state, because they were never in the pool. A survey answered only by people who chose to respond online is biased, because volunteers tend to feel more strongly than everyone else.
A sample estimate often comes with a margin of error. If 62% of a random sample prefers a later start time, with a margin of error of 4 percentage points, the plausible range for the whole school is 58% to 66%. A larger random sample shrinks the margin of error, but it cannot fix a biased sampling method.
In an experiment, the researcher assigns the treatment at random. If 80 volunteers are randomly split into a group that uses a study app and a group that does not, a difference in scores can be credited to the app. If students chose for themselves, the study is observational. Motivated students might be the ones who pick the app, so you can only say app use and scores are associated.
To scale up a sample result, use a proportion. If 18 of 60 sampled apples are bruised, that is 30%, so about 600 of a 2,000-apple shipment would be expected to be bruised.