Residual deviance 2026-10-06
The unscaled residual deviance of an exponential dispersion family compares its fitted mean with the saturated statistical model at a common dispersion parameter. Its normalization removes the factor in the log-likelihood. In Poisson regression and binomial regression, , so it is twice the saturated-minus-fitted log-likelihood. Under suitable regularity, has approximately a chi-squared distribution with the residual degrees of freedom; this is not an automatic accurate approximation for sparse responses. Nested deviance reductions with estimated dispersion lead to approximate F-tests after scaling by the dispersion estimate.
Each Cartesian Brownian increment has variance . Summing their normalized squares gives a chi-squared distribution with degrees of freedom. Hence is unbiased and has relative standard deviation . Its inverse is not unbiased: for , . Thus the direct inverse-diffusivity radius estimate has appreciable finite-sample bias when only a few trajectories are observed. Measurement noise or drift adds separate systematic error.
Score test Created 2026-09-24 Updated 2026-11-05
The score test evaluates the score function at the parameter estimate constrained by the null hypothesis. The likelihood gradient records which local parameter directions would improve the fit; the Fisher information supplies its sampling scale. For a fully specified regular null , with total-sample score and total-sample information ,
has an asymptotic chi-squared distribution with degrees of freedom equal to the dimension of the tested parameter. The same sample-size convention must be used in both and .
For independent Bernoulli trials, let be the number of successes among trials. Testing with gives
The signed square root is the familiar null-standardized success-count statistic. Its two-sided large-sample calibration uses ; a small sample can instead use the exact binomial distribution.
With nuisance parameters, estimate them under the null and use the corresponding efficient score and efficient information for the tested restrictions. An unadjusted information block can give the wrong scale. Unlike the Wald test, the score test does not require an unrestricted estimate under the alternative.
Shape-two Weibull scale likelihood-ratio test Created 2026-10-31 Updated 2026-11-05
For independent Weibull distribution observations of known shape two, the likelihood ratio between any scale and the null scale increases with . Under the null, has a chi-squared distribution with degrees of freedom. The same upper-tail critical region is therefore most powerful for every larger scale, giving a uniformly most powerful size- test.
Signed normal Wald statistic Created 2026-10-06 Updated 2026-11-05
For an independent normal sample with known standard deviation , the signed Wald statistic for testing mean zero is . It has the standard normal distribution under the null hypothesis and distribution under mean . Its square has a chi-squared distribution; retaining the sign permits a one-sided test.
Squared Rayleigh radius 2026-10-18
Squaring a Rayleigh distribution variable produces an exponential distribution. For ,
where the exponential distribution parameter is a rate. Equivalently, has chi-squared distribution with two degrees of freedom (statistics). An amplitude and its square therefore have different probability density functions, even though one determines the other.
This makes the Rayleigh-normal scale mixture transparent: at unit scale, is exponential, allowing a random normal scale to produce a Laplace distribution. Another equivalent parametrization uses a Weibull distribution with shape two and scale . Conventions that call this latter scale the Rayleigh scale differ by a factor of ; check the probability density function before comparing numerical parameter values.[ref]
Compute observed-minus-expected event contributions within each stratum's risk set, then add them across independent strata. Under the null, the corresponding conditional variances add. Standardizing gives an approximate normal distribution test, or its square gives a one-degree-of-freedom chi-squared distribution test. The comparison requires the usual exchangeability and independent censoring assumptions within strata.
Suppose , , are independent and identically distributed with normal distribution , where the mean and standard deviation are unknown. Define the sample mean and sample variance by
Under these assumptions,
Substitution gives the exact ratio
The unknown cancels. The denominator estimates the standard error of the sample mean; its sampling variability is precisely why the Student's t-distribution has wider tails than the standard normal distribution.
The crucial independence has a geometric explanation. Apply an orthogonal transformation to the standardized observations , choosing its first coordinate along . That coordinate is ; the remaining coordinates are independent standard normal random variables whose squared sum is . Their joint probability density factorizes because the transformation preserves the squared length. This also explains the loss of one statistical degree of freedom when estimating the mean. Uncorrelatedness alone would not establish the required independence for arbitrary data.
The result is exact for every under the stated normal distribution model. A small sample size by itself does not justify it: a sample from an arbitrary probability distribution need not have the displayed chi-squared distribution or the necessary independence.
Student's t-distribution Created 2026-09-29 Updated 2026-10-20
The Student's t-distribution describes a standard normal random variable divided by an independent estimate of its scale. For any real , let have the standard normal distribution and let have the chi-squared distribution, with and independent. Then
The parameter is called the number of statistical degrees of freedom. It need not be an integer: the chi-squared distribution is a gamma distribution with shape and rate . The independence assumption is essential; specifying only the two marginal distributions does not determine the probability distribution of their ratio.
Its probability density function on the whole real line is
where is the gamma function. Every member is a symmetric distribution about zero. At it is the Standard Cauchy distribution; as it approaches the standard normal distribution. The essential shape change is the tail: at any fixed finite , decays like a constant times , much more slowly than the standard normal density's exponential decay. Thus it is a heavy-tailed distribution. [ref]
The left panel shows the centers; the right panel uses a logarithmic scale to reveal the tails that a linear plot hides. All four curves have location zero and scale one. They do not have the same variance: the curve has no finite mean or variance, whereas has variance 3 and has variance .
More generally, the absolute moment for is finite exactly when , as follows by comparing the tail integrand with . Consequently
For there is no finite mean; symmetry alone cannot supply one. For the mean exists but the variance is infinite. For even the centered definition of variance is unavailable because the mean does not exist.
A shifted and scaled version , , has probability density function . The scale is not its standard deviation: for ,
For example, and scale give standard deviation , rather than 2. The distinction matters when using a Student's t-distribution to model data with occasional extreme values, as well as when interpreting a Student t confidence interval.
For a Gaussian linear model with an intercept and one explanatory variable, and , the regression intercept estimator is normal with variance . The residual variance estimate is independent of the coefficient estimators and has a chi-squared distribution with degrees of freedom. Consequently the displayed statistic has a Student t-distribution under , and a two-sided level- test rejects for .
Uncertainty in estimated inverse-variance weights arises when the variances used in inverse-variance weighting are themselves estimated. A very small estimated variance can give one observation or group a large random weight. The familiar variance formula for fixed, known weights cannot simply be reused with estimated variances treated as exact. NIST specifically warns that weights estimated from few replicates can substantially and unpredictably affect a weighted least squares analysis.[ref]
Suppose independent groups contain independent normal random variables with common mean and group variance . Let and be the group sample mean and unbiased sample variance, and form
For normal samples, each sample mean is independent of its sample variance. Conditional on the estimated weights, the estimator remains unbiased and has
Thus the law of total variance gives unconditional variance equal to the expectation of this expression. The true still appear; replacing them by the same random does not establish the actual conditional variance. Outside the independence assumptions, even conditional unbiasedness can fail.
An original two-group example takes and true , so each group mean has variance . If observed sample variances are and , the estimated normalized weights are and . For group means and , the combined estimate is . The plug-in variance is
but the true conditional variance under the stipulated model is
The corresponding standard deviations are about and . Equal weights would have variance . An accidentally small estimated variance has both overstated precision and moved the combination away from the optimal equal weights in this example.
Reciprocal moments of an estimated variance explain the unstable precision estimate. For a normal sample with , has a chi-squared distribution. Integrating its density gives
Consequently for , while the second moment requires . The density and its gamma-distribution parameterization are documented in the primary SciPy implementation notes.[ref] With , the inverse sample variance has a finite mean twice the true precision and an infinite variance; the normalized weights remain bounded, so this does not imply their variance is infinite.
More replication or a justified smooth variance model can stabilize weights. NIST describes direct replicate-based weight estimation and variance-function modeling as distinct approaches.[ref] A proposed validation repeats the complete weight-estimation and combination procedure on simulated or replicated groups, checking actual interval coverage as well as point estimates. A bootstrap must reproduce the weight-estimation stage; holding the fitted weights fixed tests a different procedure.
Uncertain variance in heterodyne lognormal bias correction matters when a measured overlap is corrected using an estimated, rather than known, logarithmic variance. The known-variance correction is a nonlinear function of that estimate, so replacing the variance by an unbiased estimate does not automatically preserve unbiasedness of the corrected overlap.
Suppose , with , and use . If the variance estimate is independent of the overlap readout,
Even if , Jensen's inequality gives . Consequently the mean corrected estimate is at least under these independence and unbiased-variance assumptions. Equality requires a nonrandom variance estimate in the ordinary finite-variance case.
An original exact example uses a variance estimate from an independent normal calibration sample, with following a chi-squared distribution with degrees of freedom. Its Laplace transform gives
The logarithm of this factor is , which is positive for . With stipulated and , the factor is : about a 9.91% remaining mean shift. This intentionally broad statistical example is not a typical receiver specification.
For fixed and large , the logarithm is . More independent calibration information reduces this effect. Merely taking more overlap readings while retaining the same uncertain variance estimate does not remove its influence.
Dependent variance correction in heterodyne normalization requires the joint quantity instead of the product of expectations above. A variance inferred from the same readout cannot simply be declared independent. Propagate the joint calibration and readout model, with a stated target summary. JCGM 101 treats input distributions from prior uncertainty calculations and numerical propagation explicitly.[ref]
The optical gain-matching condition for heterodyne normalization still concerns physical calibration paths. This calculation addresses the statistical correction of residual log-errors; it does not imply that an imperfect RF or DC calibration can be repaired solely by subtracting a fitted variance.
Wilson-Hilferty transformation Created 2026-10-20 Updated 2026-11-05
For , the Wilson-Hilferty transformation uses the approximation
The cube root reduces the right skewness of the chi-squared distribution. Inverting gives the approximate -quantile , where is a standard normal quantile. This is an approximation, especially imperfect for extreme tails or small degrees of freedom. For a Gamma distribution with shape and rate , apply it to , which has .[ref]
Yates's correction for continuity Created 2026-10-14 Updated 2026-11-05
For a two-by-two contingency table, subtracting one half from each absolute observed-minus-expected count difference, with truncation at zero, gives the displayed corrected Pearson chi-squared statistic. Expected counts use the independence model . The subtraction compensates approximately for using a continuous chi-squared distribution for discrete counts. It can make the test conservative. The corrected statistic should be distinguished from the ordinary uncorrected Pearson chi-squared statistic.