Analysis of variance for a Latin square Created 2026-10-14 Updated 2026-10-25
For one response per cell in an order- Latin square, fit an intercept and additive row, column and treatment effects with sum-to-zero constraints. Each factor contributes statistical degrees of freedom; the residual has . The centered factor spaces are mutually orthogonal, so sequential sums of squares do not depend on fitting order. With an additional orthogonal symbol factor from a Graeco-Latin square, residual statistical degrees of freedom become . Exact F-tests require independent equal-variance normal errors; unmodeled interactions can invalidate the error term.
Bernard Lewis Welch Created 2026-10-25 Updated 2026-11-05
Bernard Lewis Welch developed inference for means when sample variances cannot be treated as estimates of one shared population variance. His 1947 paper explicitly generalized Student's problem to several different population variances.[ref]
The familiar two-group Welch's t-test estimates the standard error of a mean difference using , rather than pooling both sample variances under an equal-variance model. Its reference distribution uses approximate statistical degrees of freedom. This makes the distinction between unequal population variances and unequal sample sizes explicit: both enter the uncertainty calculation, and unequal group sizes do not by themselves justify assuming equal variances.
Beta sampling by gamma ratios Created 2026-10-14 Updated 2026-11-05
independent gamma distributions and with a common rate yield a Beta distribution . The change of variables , has Jacobian determinant , and factors the joint probability density function into a Beta distribution density in and a gamma distribution density in . In particular, independent chi-squared distributions with and degrees of freedom give the required ratio.
For independent groups each containing independent Bernoulli trials with common probability , the group count has a binomial distribution. If groups have count , the maximum-likelihood estimate is . Form the Pearson chi-squared statistic using . With all categories retained, an interior true parameter, and sufficiently large expected counts, its null limit is a chi-squared distribution with statistical degrees of freedom: one degree is removed by the fixed total and another by fitting . Small expected counts require care with this approximation. If categories are pooled, the fitted parameter and calibration should correspond to the pooled statistical model; simply pooling a statistic after fitting the unpooled model does not automatically give the usual chi-squared limit.
Chi-squared distribution Created 2026-09-24 Updated 2026-11-05
A Chi-squared distribution with statistical degrees of freedom is a Gamma distribution with shape and rate . Its Probability density function is
For a positive integer , a variable is the sum of squares of independent standard normal random variables. The gamma definition extends the law to noninteger positive degrees of freedom; it does not require a fractional number of Gaussian coordinates.
Its mean is , its variance is , and its Moment-generating function is for . These follow directly from the Gamma density or by multiplying the squared-normal transforms in the integer case.
The law appears when Gaussian deviations are aggregated as a squared length. For iid observations with and , the Sample variance satisfies . Centering removes the one-dimensional sample-mean direction, leaving independent Gaussian coordinates after an orthogonal change of basis. This exact normal-sampling law underlies confidence intervals for a variance; it is not a universal variance-sampling formula for arbitrary data.
Chi-squared stochastic ordering Created 2026-09-24 Updated 2026-11-05
Chi-squared distributions are stochastically increasing in their statistical degrees of freedom. For any real , take independent and . Gamma additivity gives
For integer , this is the familiar construction by adding extra independent squared standard normal variables. For noninteger degrees of freedom, multiplying the gamma moment-generating functions proves the same statement.
Under this coupling the larger-degree variable is larger on every sample path. Hence for every , and in fact the inequality is strict for : average and use and the strictly positive chi-squared density. Therefore every quantile at a fixed probability in strictly increases with the degrees of freedom. Comparing chi-squared critical values must account for this change; the distribution is not fixed while only its label changes.
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 satisfy
Its 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.
Deviance goodness-of-fit test Created 2026-10-06 Updated 2026-11-05
A deviance goodness-of-fit test compares a fitted generalized linear model with its saturated statistical model. With independent response groups, an identifiable rank- mean model and known common dispersion parameter , the candidate statistic is the scaled deviance
In an appropriate large-information regime, compare it with a chi-squared distribution on statistical degrees of freedom. For a normal linear model this is exact when is known: is the residual sum of squares. For Poisson regression, ; adequate expected cell counts are still needed for a useful approximation.[ref]
More individual Bernoulli observations do not automatically justify comparing their residual deviance with . Small cell counts, dependence and overdispersion can invalidate that calibration. An unknown dispersion parameter also needs an explicit estimation and testing convention; estimating it from and then treating as an independent goodness-of-fit check is circular.
Drift-aware homogeneity analysis estimates material variation while accounting for time-dependent changes in the measurement system. Its purpose is to prevent instrumental drift from being assigned to differences between containers in a homogeneity study with replicated test portions. The recorded order is part of the experimental data.
A simple measurement model is
with container contribution , linear drift rate and residual . A linear regression can treat container differences as fixed effects; a random effect formulation instead models a between-container variance. Either approach must distinguish the time term from the container terms. Homogeneity-study time-term identifiability depends on the actual sampling schedule.
An original noiseless example uses A at times 0 and 5 with readings 100 and 105, B at 1 and 4 with readings 103 and 106, and C at 2 and 3 with readings 101 and 102. Each within-container difference divided by its time separation gives drift rate one. Subtracting that rate times observation time yields A readings 100, B readings 102 and C readings 99. The adjustment retains the true stipulated container differences instead of forcing all means to coincide.
With noise present, the rate is uncertain. Substituting a fitted rate and then analyzing the adjusted observations as if that rate were known omits its uncertainty and induced covariance. For example, six observations with an intercept, two independent container indicators and one time coefficient leave only two residual statistical degrees of freedom when the design has full rank. A more elaborate drift curve consumes information and may explain away genuine material differences.
Randomized order and control observations help diagnose the model. The IUPAC protocol requires randomized analysis of duplicate portions and checking trends or other suspect data features.[ref] Separately, NIST's dimensional-comparison designs deliberately balance some drift contributions through observation order.[ref]
Plot residuals against time and container, retain independent controls, and state how the time dependence was chosen. A fitted linear term does not establish that all drift is linear. If time and container effects cannot be distinguished, the study needs additional information or a better design rather than an unjustifiably precise homogeneity estimate.
Edwin Bidwell Wilson Created 2026-10-24 Updated 2026-11-05
Edwin Bidwell Wilson was a mathematician and statistician, remembered in statistics for the Wilson score interval. His 1931 paper with Margaret M. Hilferty introduced an influential approximation to the chi-squared distribution. His firsthand National Academy of Sciences memoir lists the paper under his full name.[ref]
Their Wilson-Hilferty transformation uses for a chi-squared distribution with statistical degrees of freedom. Its approximate mean and variance give a normal approximation useful for converting quantiles between the two distributions.[ref]
F-distribution Created 2026-09-29 Updated 2026-11-05
The F-distribution with degrees of freedom is the distribution of for independent chi-squared random variables and .
For independent strata, each containing an contingency table, test independence of the two table variables within every stratum. Under adequate expected counts and the usual multinomial distribution sampling assumptions, each stratum's Pearson chi-squared statistic has asymptotic chi-squared distribution with statistical degrees of freedom. Their sum has asymptotic chi-squared distribution with statistical degrees of freedom, by the additivity of independent chi-squared distributions. This is a test of conditional independence, not of equality of a common effect across strata; the latter needs an interaction contrast or another specified homogeneity test.
Let for independent and identically distributed observations, with having an exponential distribution of unknown scale . Test the null hypothesis within . For nondegenerate data, the likelihood function is
Maximizing over the scale gives , and the resulting profile likelihood increases with . Thus the unrestricted maximum-likelihood estimates are and , while the null estimate is . The likelihood ratio is consequently
Under the null, put and . The memoryless property gives independent gamma distributions of shapes and and common scale . For a direct verification, fix which observation is the minimum and write its value as and the other observations as . Their joint density is proportional to , which factors into an exponential density for the minimum and independent exponential densities for the excesses. The minimum's index does not affect the resulting laws of .
By beta sampling by gamma ratios, . Therefore
An exact test size is obtained by rejecting for . Equivalently,
Although the alternative adds one endpoint parameter, a regular one-parameter Wilks theorem calibration is incorrect: the support depends on the endpoint, and the null lies on its boundary. In particular, tends to a chi-squared distribution with two statistical degrees of freedom, not one.
Margaret M. Hilferty 2026-11-05
Margaret M. Hilferty coauthored the 1931 paper The Distribution of Chi-Square by Edwin B. Wilson and Margaret M. Hilferty. The original byline identifies her with Harvard's Department of Vital Statistics; no fuller personal name is inferred from the initials.[ref]
Their Wilson-Hilferty transformation reduces the skewness of a chi-squared distribution by taking a cube root. For , the standardized quantity is approximately standard normal. It provides a practical approximation to quantiles and tail probabilities, with accuracy depending on the statistical degrees of freedom and the probability region.[ref]
Mean square in ANOVA 2026-10-07
A mean square in ANOVA divides a sum of squares in ANOVA by its positive number of statistical degrees of freedom. If the corresponding residual subspace has scalar error covariance , its expectation is . Its scale depends on whether raw observations or group means were projected.
Multinomial central limit theorem Created 2026-10-07 Updated 2026-11-05
The vector of category counts is the sum of independent categorical indicator vectors. Their means are and covariance is , so the multivariate central limit theorem gives the displayed limit. At uniform probabilities on categories, Pearson-standardized deviations have covariance equal to the orthogonal projection onto the -dimensional subspace of zero-sum vectors. Their squared norm has an asymptotic chi-squared distribution with degrees of freedom.
Multiple-imputation confidence interval Created 2026-10-06 Updated 2026-11-05
Use the pooled estimate and total variance from Rubin's rules. A basic large-complete-sample scalar interval is , with for . Small complete-data samples need an appropriate statistical-degrees-of-freedom adjustment; the normal limit applies as the available degrees of freedom become large.
Noncentral chi-squared distribution Created 2026-10-06 Updated 2026-11-05
If are independent standard normal random variables and , then has a Noncentral chi-squared distribution with degrees of freedom and noncentrality . For it reduces to the chi-squared distribution. Squaring a shifted signed normal Wald statistic gives the case .
Noncentral t-distribution Created 2026-10-05 Updated 2026-11-05
If and are independent, then has the noncentral t-distribution with statistical degrees of freedom and noncentrality . It describes the test statistic of a Student's t-test under a shifted alternative hypothesis. At it reduces to the Student's t-distribution. This is different from adding a constant to a central Student t-distribution: here the shift is inside the random denominator. For a normal distribution sample tested against null hypothesis , the noncentrality is .