PASL cheat sheet

A blueprint assigns each content domain a percentage of scored items; the domain percentages sum to 100% of the scored exam.

Only operational (scored) items contribute to a reported score; embedded pretest items are unscored and do not affect the result.

Criterion-referenced exams compare a candidate against a fixed performance standard, whereas norm-referenced exams compare candidates against each other.

The cut score is set before administration using standard-setting procedures such as a modified Angoff panel, not by failing a fixed percentage of candidates.

Scaled scores convert raw scores from different forms onto a common metric so that forms of unequal difficulty can be compared.

Standard error of measurement defines a confidence band around an observed score; a score within one SEM of the cut is inside measurement error.

Item difficulty (p-value) = number answering correctly ÷ number attempting the item; values near 1.0 denote easy items.

Item discrimination (point-biserial) = correlation between item score and total score; negative values usually flag a miskeyed or flawed item.

In a single-best-answer item, exactly one option is defensibly correct; every distractor must be plausible yet unambiguously wrong.

For negatively worded stems (asking which option is NOT correct), restate the stem positively before evaluating the options.

Absolute qualifiers such as 'always', 'never', and 'only' make an option too extreme to be correct in most professional-judgment items.

When two options both appear correct, the differentiator is usually a qualifier or condition in the stem rather than the general principle.

For 'first action' items, select the earliest step in the accepted sequence, not the most complete or definitive intervention.

For 'most appropriate' items, select the option that fits the stated constraints; the ideal response with unlimited resources is a distractor.

Before any calculation item, convert all quantities to one unit system; answers differing by orders of magnitude usually indicate a unit error.

Flag uncertain items and return to them; no single item should consume a disproportionate share of a timed exam.

Valid informed consent requires four elements: capacity, disclosure of risks and alternatives, voluntariness, and documentation.

Confidentiality may be breached only with consent, under a legal requirement, or to prevent serious imminent harm.

Mandatory reporting duties are statutory and override ordinary confidentiality when their trigger conditions are met.

A professional performs only tasks within legal scope and demonstrated competence; anything outside is referred or delegated per regulation.

A conflict of interest is managed by disclosure first and recusal when the conflict cannot otherwise be neutralized.

Records should be contemporaneous, objective, and factual; errors are corrected by a dated, signed addendum, never by alteration or deletion.

Dual relationships (client plus friend, relative, or business partner) are avoided where possible; when unavoidable, safeguards and boundaries are documented.

Sensitivity = true positives ÷ (true positives + false negatives); it is the proportion of affected people correctly identified.

Specificity = true negatives ÷ (true negatives + false positives); it is the proportion of unaffected people correctly identified.

Positive predictive value = true positives ÷ (true positives + false positives); unlike sensitivity, it changes with the prevalence of the condition.

Risk ratio is computed in cohort studies (incidence exposed ÷ incidence unexposed); odds ratio is used in case-control studies.

Median is preferred over mean when a distribution has extreme outliers, because the median is resistant to their influence.

In a normal distribution, roughly 68% of values lie within one standard deviation of the mean and about 95% within two.

A 95% confidence interval that excludes the null value (0 for a difference, 1 for a ratio) corresponds to statistical significance at alpha = 0.05.

Percent change = ((new value − old value) ÷ old value) × 100; a correlation between two variables does not by itself establish causation.