Frameworks
Five tools that work on any statistics problem. Use the first one when you freeze: answer three questions and it hands you the formula.
1. Which formula? Answer the questions
What is the question about?
2. The three moves in every probability question
- OrAdd them, then subtract the overlap. Heart or face card = 13 + 12 − 3.
- AndMultiply. Update the second count if there is no replacement.
- GivenShrink the list to what you were told, then count inside it. The given goes on the bottom.
3. Four S for any word problem
- 1SketchDraw it: a bell with the cutoff, a tree, a two-way table. Ten seconds, and half the mistakes disappear.
- 2Set upWrite the formula with symbols before numbers. Name μ, σ, n, p. Decide sample or population.
- 3SolvePlug in. Say the ballpark out loud first, then use the calculator.
- 4Sanity checkIs it between 0 and 1? Is the SD smaller than the range? Is the average less noisy than one value? Put units on the answer.
4. The study loop
- 1LearnOne line and an example.
- 2LabDrag the control until the number makes sense.
- 3QuizAnswer before you reveal.
- 4DrillFresh numbers, on a clock.
- 5Due todayThe misses come back.
This is spaced retrieval practice with interleaving, the two study techniques with the strongest evidence (Dunlosky et al., 2013). Ten minutes a day through the loop beats an hour of rereading. Your topics on Practice only fill in from drills, on purpose.
5. Reading a question: the words that decide everything
If it says: at least one1 − P(none)Go to Probability, binomial
If it says: claim, significant, evidence that, test at 5%Hypothesis test: (estimate − H₀) / SE, then the p-valueGo to Hypothesis tests
If it says: given, of those, amongConditional: shrink the worldGo to Probability
If it says: average of n, sample mean, total of nStandard error σ/√n, or σ√n for a totalGo to CLT
If it says: per minute, per hour, arrivals, occurrencesPoisson; rescale λ to the windowGo to Poisson
If it says: n trials, each a success or failure, how manyBinomialGo to Binomial
If it says: within ± E, margin, how many to sampleSample size, round upGo to Confidence intervals
If it says: length of the intervalE is half the lengthGo to Confidence intervals
If it says: s instead of σ, small nt, not zGo to Confidence intervals
If it says: holding the others fixedA multiple regression coefficientGo to Multiple regression
6. Why the page is built this way
Plain sentence, then example, then formulaConcrete before abstractCognitive load research (Sweller)
A picture next to every ideaDual codingPaivio; Mayer's multimedia principles
Solved problems with the answer highlightedWorked examplesSweller's worked-example effect
Quiz mode and drillsRetrieval practiceRoediger and Karpicke; rated high utility in Dunlosky et al. (2013)
Missed questions come back laterSpaced practiceCepeda et al.; Dunlosky et al. (2013)
Drills mix topicsInterleavingRohrer and Taylor
Desk and court analogiesElaborative encodingDunlosky et al. (2013)
Timed sets pitched just above your levelFlowCsikszentmihalyi; self-determination theory (Deci and Ryan)