Scio NSZ OSP cheat sheet

Analogy (A : B :: C : D) requires one fixed relation; test the relation in both directions and reject any option that needs two different relations.

Relation types to name explicitly: part–whole, cause–effect, function, degree/intensity, category–member, synonym, antonym, sequence, and producer–product.

Antonym/synonym items: the correct answer must match the word's part of speech and register; a near-synonym with a different grammatical form is a distractor.

Word-in-context: substitute each option into the sentence and keep only the one that preserves both literal meaning and the sentence's logical direction.

Odd-one-out: find the property shared by all but one item; the shared property must be a single criterion, not a list of unrelated similarities.

Sentence completion with two blanks: check each option against both blanks simultaneously; an option fitting only one blank is always wrong.

Passage inference: a valid conclusion must follow from the text alone, so any statement requiring outside knowledge or probability is out of scope.

Scope words decide correctness: all, none, some, only, always, never, at least, at most - compare the quantifier in the answer with the quantifier in the text.

Percent change = (new − old) / old × 100; a rise then fall of the same p% never returns to the original value (net factor = (1+p)(1−p) = 1 − p²).

Successive percentage changes multiply: +20% then −15% gives 1.20 × 0.85 = 1.02, i.e. +2%.

Percentage points vs percent: a move from 20% to 25% is +5 percentage points but +25% relative.

Ratio scaling: if a : b = 3 : 5, the parts are 3k and 5k; the whole is 8k, so one part equals its share of the total divided by the sum of shares.

Averages: total = average × count; if two groups have sizes n₁, n₂ and means m₁, m₂, the combined mean is (n₁m₁ + n₂m₂)/(n₁ + n₂), weighted by group size.

Work/rate: combined rate = sum of individual rates (1/a + 1/b); time together = ab/(a + b) for two workers.

Speed: distance = speed × time; average speed over equal distances is the harmonic mean 2v₁v₂/(v₁ + v₂), not the arithmetic mean.

Simple vs compound interest: simple = P·r·t; compound = P(1 + r)^t − P, so compound exceeds simple for t > 1 with r > 0.

Number-series items: first check differences, then ratios, then alternating/interleaved subsequences, then squares/cubes or ±n patterns.

Ordering/sequencing puzzles: convert each clue into an inequality (A before B → A < B) and chain transitive relations before testing options.

Grouping/assignment puzzles: track fixed memberships and exclusions separately; a clue 'X and Y are never together' forbids exactly one pairing, not the whole group.

If–then clues are one-directional: from 'if P then Q' you may infer not-Q → not-P (contrapositive), never Q → P.

'Only if' reverses the arrow: 'P only if Q' means P → Q, so Q is necessary and P is sufficient.

'Unless' means 'if not': 'P unless Q' = not-Q → P, equivalently P or Q.

Elimination strategy: apply the most restrictive clue first; each option that violates any single clue is discarded immediately.

Counting arrangements: n distinct items in a row = n!; choose and order k of n = n!/(n − k)!; unordered selection = n!/(k!(n − k)!).

Spatial/table puzzles: draw a grid with one row per position or person and mark ✓/✗; a filled row or column forces the remaining entries.

Argument core = premises (evidence) + conclusion (claim); the conclusion is what the author wants accepted, and everything else supports it.

Assumption questions: the right answer is necessary (negate it and the argument collapses) and is not restated in the text.

Strengthen/weaken items: the correct option changes the link between evidence and conclusion; irrelevant background facts never do.

Correlation is not causation: alternative explanations are reversed causation, a common cause, or coincidence/selection bias.

Flaw types to name: hasty generalization, false dilemma, circular reasoning, ad hominem, straw man, appeal to authority, and confusing necessary with sufficient conditions.

Sampling: a valid generalization needs a sample that is random and representative; a larger biased sample stays biased.

Necessary vs sufficient: 'A is necessary for B' = B → A; 'A is sufficient for B' = A → B; mistaking one for the other is a standard trap.

Quantifier traps: 'some' means at least one, not most; 'most' means more than half, and neither implies 'all'.

Study prerequisites: spaced repetition and self-testing (retrieval practice) produce better retention than rereading; plan timed practice under exam conditions.