INRAT cheat sheet
INRAT · Keep the key ideas within reach.
Domain 1 - placeholder (official INRAT outline not supplied; content pending source verification): quantitative methods and study-design fundamentals
A p-value is the probability, computed under the null hypothesis, of a test statistic at least as extreme as the one observed. It is not the probability that the null hypothesis is true, nor the probability that results are due to chance.
Power = 1 − β. For a two-sided two-sample test, required n per group ≈ 2σ²(z_{1−α/2} + z_{1−β})² / Δ², where Δ is the smallest difference worth detecting; n grows with σ² and shrinks with Δ².
Type I error (α) is rejecting a true null; Type II error (β) is not rejecting a false null. With n and effect size fixed, lowering α reduces power and raises β.
A 95% confidence interval is a procedure: in repeated samples, 95% of intervals so constructed contain the true parameter. Once a single interval is computed, the frequentist interpretation attaches to the procedure, not to that interval.
Sensitivity = TP/(TP+FN); specificity = TN/(TN+FP); PPV = Se·p / [Se·p + (1−Sp)(1−p)], where p is prevalence. PPV and NPV change with prevalence while sensitivity and specificity do not.
In a randomized complete block design every block contains each treatment exactly once, treatment order is randomized within blocks, and the model includes a block term that removes between-block variability from the residual error.
A split-plot design has two error strata: whole-plot error tests the main-plot factor, and subplot error tests the subplot factor and its interaction. Using the wrong error term invalidates the F-test.
The Pearson correlation r measures only the strength and direction of linear association and is unit-free; r² gives the proportion of variance in one variable linearly accounted for by the other. Neither establishes causation without design-based identification.
Measurement scales determine permissible statistics: nominal supports frequencies and modes; ordinal supports medians and rank tests; interval supports means and differences but not ratios; ratio supports all arithmetic including ratios, because it has a true zero.