Alternative hypothesis Created 2026-10-14 Updated 2026-11-05
The alternative hypothesis specifies the parameter values or distributions against which a null hypothesis is tested. If the null is , the alternative is with . For example an upper one-sided comparison uses , while a two-sided comparison uses . The alternative determines the relevant rejection tail. A p-value is computed under the null; it is not the probability that the alternative is false.
Consider a two-sided test rejecting when , with having a normal distribution of mean and variance one. Here is the cumulative distribution function of the standard normal distribution. Under a nonzero alternative , the type II error probability is the displayed function. It tends to the nonrejection probability under the null as , and to zero as . A particular observed has positive density under all these alternatives. Its null-calibrated p-value therefore does not select one value of . The power function describes repeated-sampling performance at specified alternatives.
Benjamini-Hochberg leave-one-out identity 2026-10-05
Let be the number of rejections by the Benjamini-Hochberg procedure at level , and let be the rejection count after replacing by zero. Then andIf was rejected, lowering its p-value does not change any ordered p-value beyond rank , so . Conversely, if , the other rejected values together with satisfy the original threshold; monotonicity gives . For a valid true-null p-value independent of the others, conditioning on bounds the expectation of the right side by . Summing over true nulls proves false discovery rate at most .
Benjamini-Hochberg leave-two-out identity 2026-10-14
Replace two p-values by zeros in the Benjamini-Hochberg procedure at level , keeping the original denominator , and let be the resulting rejection count. Equivalently, remove the two values and apply the step-up critical values to the remaining ordered p-values; their rejection count is . For ,If both original values are rejected, lowering them changes no ordered rank greater than , so the rejection count stays . Conversely, if the modified count is and both original values meet the rank- threshold, reinserting them leaves at least qualifying values. Monotonicity under lowering gives at most original rejections. This identity is deterministic and does not require independence or a distributional assumption.
Benjamini-Hochberg procedure Created 2026-09-24 Updated 2026-11-05
Order p-values as and take the largest with . The Benjamini-Hochberg procedure rejects the hypotheses corresponding to and controls the false discovery rate under independence and several forms of positive dependence.
If two independent approximately Gaussian mean lifetime estimates have known standard deviations , equality of their true lifetimes implies that their difference has mean zero and variance . Its squared standardized difference therefore has a one-degree-of-freedom chi-squared distribution, giving the statistic in the title. A chosen significance threshold converts its upper-tail P-value into a compatibility decision.
Testing a fitted decay curve against a fixed numerical lifetime is a different question: that test does not include the external measurement uncertainty. Comparing the two experiments requires both uncertainties or a joint likelihood with a common lifetime. Correlated estimates replace the denominator by .
Confounder Created 2026-09-24 Updated 2026-11-05
A confounder is a pretreatment variable relevant to removing confounding for a specified treatment-outcome question. A common cause with arrows is a basic example: it creates a backdoor path between treatment and outcome .
Association with both variables is not a sufficient definition. In an appropriate causal directed acyclic graph, conditioning on a common effect can instead introduce collider bias. A confounder is a causal adjustment role, rather than any predictor with a small p-value.
One precise definition identifies a confounder as a member of some minimal sufficient adjustment set: there is a set containing that makes the relevant potential outcomes independent of treatment, and no proper subset of that set does so. The role can therefore depend on which other covariates are available and on the effect being estimated. A variable that merely reduces residual bias need not satisfy this definition.[ref]
On the definition of a confounder by Tyler J. VanderWeele and Ilya Shpitser. Sections 1–3 and concluding remarks: association-only criteria and the minimally sufficient adjustment-set definition
For one two-sided t procedure with observed estimate , positive standard error , null value and reference Student t-distribution with statistical degrees of freedom, the test statistic and p-value must satisfyIts symmetric confidence interval at coverage probability is and therefore has midpoint . When the same standard error and reference distribution are used, excluding is equivalent to and hence for a continuous reference distribution. This proves the test–interval equivalence and supplies several independent checks on a printed output. An approximate Welch's t-test uses its same approximate statistical degrees of freedom throughout. Rounding can blur a boundary decision but cannot explain a confidence-interval midpoint far from the reported estimate.
Contingency coefficient 2026-10-25
The contingency coefficient transforms the Pearson chi-squared statistic as . It is an unsigned measure of statistical association, not a p-value or a direction of an ordinal relationship. Writing gives . Unlike Cramér's V, its attainable maximum depends on table dimensions, so comparisons between differently sized tables need care.
Cramér's V 2026-10-25
For an contingency table with observations and positive row and column totals, Cramér's V rescales the Pearson chi-squared statistic to a dimensionless measure of statistical association. It lies between zero and one and is unchanged by reordering category labels. It therefore describes the magnitude of dependence, without an ordinal direction. A small p-value can accompany a small in a large sample: evidence against independence and the magnitude of the departure answer different questions. The displayed quantity is the usual uncorrected sample measure.
Invert a family of hypothesis tests by including each parameter value in the confidence set exactly when its test does not reject. For any true value , the events and are identical. Thus size at most gives coverage at least , and exact size gives exact coverage. When the accepted set is an interval, this is a confidence interval.
For the normal sample mean with unknown variance, the central Student t-distribution interval excludes zero exactly when exceeds the same two-sided critical quantile. A two-sided p-value strictly below therefore implies exclusion of zero from the matching closed interval. Boundary conventions agree up to events of probability zero in this continuous model.
Effect size Created 2026-10-14 Updated 2026-10-18
An effect size measures the magnitude of an observed or modeled relationship. A standardized mean difference expresses a mean contrast in standard-deviation units; other effect sizes include a correlation coefficient or odds ratio. Magnitude is distinct from the evidence against a null hypothesis supplied by a p-value: a small effect size can have a small p-value in a sufficiently large sample.
Elaboration model 2026-10-25
The elaboration model compares a pooled statistical association with the associations conditional on a third variable. Replication describes a similar pattern within the strata. Specification describes a relationship whose size or direction varies between strata. Attenuation can suggest explanation by an antecedent confounder, or interpretation through an intervening causal mediator. These causal readings require justified temporal and causal assumptions; conditional tables alone do not identify the third variable's role.[ref]
The Practice of Social Research (13th edition) by Earl Babbie. Chapter 15, printed p. 448, Table 15-5: relationship patterns and antecedent/intervening test variables
Assess the patterns through conditional probabilities or comparable effect sizes. Different p-values also reflect different sample sizes and standard errors, so significant and nonsignificant results need not differ significantly.
Consider independent binomial counts under the common-probability null hypothesis, with fixed positive row sizes . Let , and assume . Conditional on , the probability mass function is the displayed multivariate hypergeometric distribution: in the joint likelihood, the factor is common to all admissible tables and cancels. The Vandermonde identity gives the denominator, so the conditional distribution is free of the unknown nuisance parameter .
For a table , its Pearson chi-squared statistic isAn exact conditional p-value sums the displayed probabilities over all tables with the same margins and . The ordering is by the Pearson chi-squared statistic, rather than by the probability of an individual table; different exact-test orderings can give different answers. Enumeration is practical when and the number of rows are small. Monte Carlo tests can approximate the same tail by sampling tables with fixed margins.[ref]
Pearson's Chi-squared Test for Count Data in the R documentation. Details: simulated contingency-table tests sample tables with fixed marginals and use the Pearson statistic
An exact arithmetic comparison is easy too. Let be the least common multiple of the row sizes. Compare the integersince every table has the same positive denominator . This includes boundary ties without a floating-point tolerance. Conditional p-values are discrete and can be conservative; “exact” concerns their null calibration, not equality of rejection probability with every chosen significance level.
Exact false discovery rate under independent null p-values Created 2026-10-07 Updated 2026-10-14
For the Benjamini-Hochberg procedure, replace a true-null p-value by zero and call the new rejection count . The Benjamini-Hochberg leave-one-out identity identifies rejection with with the event and . If is uniform and independent of all the other p-values, each possible rejection count contributes to its expected false-discovery fraction. Summation gives per true null and the displayed exact false discovery rate. Super-uniform independent nulls yield the corresponding upper bound.
Exact null distribution of a rank sum Created 2026-10-14 Updated 2026-11-05
Under the continuous common-distribution null of the Wilcoxon rank sum test, pooled labels are exchangeable. The product records every choice of ranks: choosing assigns rank to the first sample. Its indicated coefficient counts subsets of size and rank sum . Divide by and sum the lower tail for the one-sided P-value. Equivalently the counts satisfy with , allowing an exact integer computation.
First true null argument for Holm control 2026-10-07
If the Holm step-down procedure makes a false rejection, it must reject the first true null in the ordered list. All true nulls remain at that step, so its cutoff is at most . The union bound on their marginally valid p-values gives familywise error rate at most . This proves strong control under arbitrary dependence, and super-uniformity suffices.
Higgs boson discovery in 2012 Created 2026-10-18 Updated 2026-10-20
The ATLAS experiment and CMS experiment searched proton-proton collisions at the Large Hadron Collider at collision energies of and . The ATLAS experiment used silicon tracking, a solenoid, liquid-argon electromagnetic calorimeters and an external muon spectrometer. Its July 2012 paper reported a combined local statistical significance of standard deviations near ; this paper result is distinct from the initial 4 July announcement.[ref]
The CMS experiment used silicon tracking in a solenoid, lead-tungstate electromagnetic calorimeters and external muon chambers. Electrons combine tracks and particle showers; muons combine tracking and chamber hits; photons supply shower energies and directions. Selected and candidates, with , yielded invariant mass peaks. It checked lepton efficiencies with Z boson decays, estimated fake-lepton backgrounds from control samples with relaxed identification, and fitted the smooth diphoton background from data. Its combined local statistical significance was standard deviations near .[ref]
The ATLAS experiment compared signal-plus-background and background-only hypotheses with a profile likelihood, accounting for systematic uncertainties and correlations between decay channels. Diphoton and four-lepton candidates gave precise invariant masses; decays supplied additional evidence with poorer mass reconstruction because of escaping neutrinos. A local p-value concerns one specified mass; scanning masses requires a look-elsewhere effect correction.[ref]
Two-photon decay excludes spin one under the usual assumptions, but does not prove spin zero. A mass peak does not determine every Higgs boson coupling. Production rates, branching fractions and angular correlations require additional tests.[ref]
Hochberg procedure Created 2026-10-05 Updated 2026-10-06
Order p-values and reject the first hypotheses, where , with for an empty set. Under independent valid true-null p-values, this step-up procedure controls the familywise error rate. One proof embeds its rejections in the closed testing procedure with Simes tests: every intersection containing a rejected hypothesis passes its local Simes test, so closed-testing control of the familywise error rate applies.
Holm–Bonferroni method Created 2026-09-24 Updated 2026-11-05
The Holm–Bonferroni method orders p-values and compares the th smallest with , stopping at the first failed comparison. It controls the familywise error rate under arbitrary dependence.