INRAT flashcards
INRAT · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Null hypothesis
- A statement of no effect or no difference used as the default in significance testing; the observed data are assessed for compatibility with it.
- Type I error
- Rejecting a true null hypothesis. Its probability is α, the significance level fixed before analysis; conventional values are 0.05 and 0.01.
- Type II error
- Failing to reject a false null hypothesis; its probability is β. Statistical power equals 1 − β and is usually targeted at 0.80 or above.
- Statistical power
- The probability of detecting a true effect of a specified size at a given α and sample size; numerically equal to 1 − β.
- Confidence interval
- A range computed from data such that, under repeated sampling, a stated proportion (for example 95%) of intervals so constructed contain the true parameter.
- Standard error of mean
- The standard deviation of the sampling distribution of the sample mean; computed as s/√n and decreasing as sample size increases.
- p-value definition
- The probability, assuming the null hypothesis, of a test statistic at least as extreme as that observed; it is not the probability the null is true.
- Sensitivity
- The proportion of truly positive cases correctly identified by a test, TP/(TP+FN). Low sensitivity means the test produces more false negatives.
- Specificity
- The proportion of truly negative cases correctly classified by a test, TN/(TN+FP). Low specificity means the test produces more false positives.
- Positive predictive value
- The probability that a positive result is a true positive: (Se × prevalence) / [Se × prevalence + (1 − Sp)(1 − prevalence)].
- Negative predictive value
- The probability that a negative result is a true negative: [Sp(1 − p)] / [Sp(1 − p) + (1 − Se)p], where p is the prevalence.
- Randomized complete block
- A design in which each block contains every treatment once, with randomization inside blocks; blocks absorb a known nuisance source of variability and raise precision.
- Completely randomized design
- A design assigning all experimental units to treatments entirely at random; the analysis uses a single error term and assumes homogeneous units.
- Split-plot design
- A design with a hard-to-change whole-plot factor and an easier subplot factor; it requires two error strata and two separate F-tests.
- Replication
- Independent repetition of each treatment across experimental units; replication supplies the error estimate needed to test treatment effects.
- Randomization
- Assignment of treatments to experimental units by a chance mechanism; it protects against systematic bias arising from unmeasured confounders.
- Purpose of blocking
- Grouping similar experimental units into blocks so that between-block variability is removed from residual error, thereby increasing precision of treatment comparisons.
- Confounding variable
- A variable associated with both the exposure and the outcome that distorts the observed exposure–outcome relation unless controlled by design or analysis.
- Internal validity
- The extent to which a study estimates the causal effect it intends in the population studied, free from confounding and selection bias.
- External validity
- The extent to which study results generalize to other populations, settings, and time periods beyond those actually examined.
- Nominal scale
- A measurement scale using unordered categories or labels, such as species or sex; only frequencies and the mode are statistically meaningful.
- Ordinal scale
- A measurement scale with ordered categories whose intervals are unequal, such as severity scores; medians and rank-based tests apply.
- Interval scale
- A scale with equal differences but an arbitrary zero, such as Celsius temperature; differences and means are meaningful, while ratios are not.
- Ratio scale
- A scale with equal intervals and a true zero, such as mass in kilograms; all arithmetic, including ratios between values, is meaningful.
- Coefficient of determination
- The proportion of variance in the response explained by a linear regression on the predictor; ranges from 0 to 1.
- Pearson correlation
- A measure of the strength and direction of linear association between two continuous variables; ranges from −1 to 1 and is unit-free.
- ANOVA assumptions
- Independent observations, normally distributed residuals, and homogeneous variance across groups; violations distort type I error rates and test validity.
- Analysis of covariance
- ANOVA extended with a continuous covariate to reduce residual variance and adjust treatment means for baseline differences between groups.
- Intention-to-treat analysis
- Analyzing participants according to their randomized group regardless of adherence or dropout; it preserves randomization and avoids attrition-related bias.
- Informed consent elements
- Disclosure of purpose, risks, benefits, alternatives, confidentiality, voluntariness, and the right to withdraw at any time without penalty.
- Sample size determinants
- Effect size, variance, α, power, and design structure determine the required n; smaller effects and higher power demand larger samples.
- Multiple comparisons problem
- Testing many hypotheses inflates family-wise type I error; control it with Bonferroni (α/m), Holm step-down, or false discovery rate procedures.