Rate parameter of a gamma distribution 2026-11-05
In the shape-rate convention for a gamma distribution, the rate parameter appears in the exponential decay:It is the reciprocal of the scale parameter , not the shape . If has this distribution, then has shape and rate one. This follows by the change of variables formula for a probability density.
The mean is and the variance is . Thus doubling the rate at fixed shape halves the mean and quarters the variance. When measures time, has units of inverse time. For positive integer shape , can be the waiting time for the th event of a rate- Poisson process, so the parameter is literally its event rate. Shape-rate and shape-scale software conventions must be distinguished before substituting numerical parameters.
Write , , and . Apply the Poisson size-bias identity with , and expand using the binomial theorem and independence. A finite th severity moment suffices. In particular, the aggregate expected value, variance, and third central moment are , , and , respectively; the latter two use raw severity moments.
Rayleigh distribution Created 2026-10-06 Updated 2026-10-18
The Rayleigh distribution is the probability distribution of the Euclidean norm of a two-dimensional random vector with independent zero-mean normal random variables of equal variance as components. If are independent, thenHere is the standard deviation of each component, not of the radius . At this gives the unit-scale probability density function .[ref]
The Rayleigh Distribution. Random: Probability, Mathematical Statistics, Stochastic Processes by Kyle Siegrist; The General Rayleigh Distribution, Definition and Distribution Functions: independent equal-scale centered normal components and radial density
This geometry turns two signed errors into one nonnegative distance. For example, suppose horizontal and vertical positioning errors are independent random variables. Their means both vanish, but the distance from the intended point is usually positive. The Rayleigh distribution with gives the probability of landing inside any centered disk. These are chosen model assumptions, rather than measurements of a particular instrument.
An apparently paradoxical feature is that the most likely location per unit area is the origin, while the radial mode is . A narrow annulus farther out has more area. Near zero that growing circumference outweighs the falling planar density; farther out the exponential function wins. Distinguishing densities measured against different elements of area or length resolves the paradox.
Rayleigh radius from Gaussian components 2026-10-18
The joint probability density of isIn polar coordinates, , , and the absolute Jacobian determinant is . Thus the joint density of radius and angle factors asIntegrating over angle gives the Rayleigh distribution. The same factorization shows that the angle has uniform distribution and is independent of the radius. This is a short change of variables argument: an annulus of width has area to first order.
The assumptions are substantive. Unequal component variances give elliptical rather than circular density contours; a nonzero mean displaces the center. Even zero covariance and equal normal distributions do not suffice without independence. For an original counterexample, let and let be an independent Rademacher random variable. Put . Then has the same normal distribution as , and . Nevertheless, all points lie on or , and has densitywhich is positive at zero, unlike the Rayleigh distribution. This pair is not a Gaussian random vector.
Readout overhead in Raman exposure allocation is the detector time that consumes the acquisition budget without collecting useful probe/reference photons. For pump-on/off exposure pairs, let each frame require a fixed closed-shutter readout duration . A wall-clock budget leaves total live timeUnder the stationary-rate, independent shot noise model of exposure-time allocation for a Raman gain ratio, its minimum approximate relative variance at fixed isReadout changes the available live time, but not this optimum ratio when the overhead is fixed and independent of the on/off allocation. The live-time formula itself ceases to describe all constraints when exposures hit saturation limits or the number of pairs is variable.
The 2003 primary stimulated-Raman experiment used 97 ms exposures and 10 ms detector readout with the exposure shutter closed. The pump shutter switched state during readout, giving successive on/off measurements.[ref]
Femtosecond Broadband Stimulated Raman: A New Approach for High-Performance Vibrational Spectroscopy by David W. McCamant and others. Data Collection, exposure/readout and S1/S2 shutter sequence preceding equation 4
These reported settings give an idealized two-frame cycle of , containing live time. The paper describes the practical full cycle as approximately 220 ms, so the two-duration sum need not include every system delay.For an original comparison at fixed wall time, the 214 ms idealization has live fraction . Relative to a hypothetical zero-overhead detector at the same rates and on/off allocation, the shot-noise variance is multiplied by , and its standard deviation by about . Using 10 ms rather than 97 ms exposures with the same 10 ms readout would give a live fraction and a variance penalty of two.
Longer frames reduce the fractional overhead but cannot be extended without limit. A maximum-exposure constraint in paired Raman acquisition can arise from saturation, while drift rejection can require frequent on/off alternation. Thus minimizing readout count alone is not a complete acquisition design; the applicable frame limits and switching cadence belong in the optimization.
Receiver-latency uncertainty in the inferred input-time horizon propagates uncertainty in timing calibration into the latest input time that could have influenced a stored output record. For one channel, the horizon in input-time horizon of a delayed optical receiver record is , with last output time and effective input-to-output timing offset . HenceThis linear variance identity retains shared timing influences.[ref]
A common calibration offset can cancel when it enters both quantities. In an original model, let errors be and , with independent components of standard deviations , and . Then and . Discarding their common-offset covariance would instead give .
The shared component must come from the actual calibration model. Here is the effective timing offset relating the input and output coordinates, so an output-clock alignment error can enter both and . A physical transport-delay error is not automatically correlated with the last-sample timestamp.
For hard timing limits, propagate validated bounds rather than reinterpret standard uncertainty as a guarantee. If is known and , then . Claiming the optimistic endpoint as though it held for every allowed delay would overstate the input-time reach.
With several channels, . Channel switching under uncertainty in a combined receiver horizon makes this a nonsmooth function when the maximizing channel is uncertain. Propagate the justified joint timing set or distribution; a derivative of only one nominally best channel can miss such switching.
NIST's primary pulse calibration measured the detector, cable and oscilloscope together using a short known optical input.[ref]
Calibration Service for Low-level Pulsed-Laser Radiometers at 1.06 µm: Pulse Energy and Peak Power by Rodney W. Leonhardt. Appendix C.1–C.2, printed pp. 36–37, Figure C1
The numerical timing errors above are chosen examples, not values from that apparatus. A timing horizon remains a necessary information limit; it does not guarantee invertibility of the receiver or bound future optical illumination.Reciprocal-estimation bias in randomized atom inclusion quantifies the nonlinear plug-in loss beyond the sign supplied by Jensen's inequality. Keep the independent-calibration model of atom-probability error in a randomized endpoint rule: true atom probability , unbiased estimate , and included probability , with the upper endpoint calibrated to actual probability . Assume , , and .
The exact algebraic identitygivesThus the loss depends on squared errors weighted by the reciprocal estimated probability. The second-order delta method approximation is , but the exact remainder shows why small estimated probabilities deserve particular attention.
If almost surely for , a nonasymptotic bound isFor an original small-error example, let , let measurement calibration return or equally often, and take , . Here and the variance is . The exact included probability is about ; the second-order approximation is , while the bounded-error calculation puts the coverage loss between about and .
The bounded-relative-error assumption can fail for a finite binomial measurement calibration: an empirical atom fraction can be zero with positive probability. Such near-zero probability estimates in reciprocal plug-in rules require an explicitly defined guard or different measurement calibration model, not division by zero. The auxiliary endpoint decision must remain independent of the future duration and measurement calibration as stipulated. The documented uniform random-variates construction supplies that auxiliary randomness; it does not validate the atom estimate.[ref]
SciPy 1.16.2 documentation. scipy.stats.uniform, support and rvs
Reciprocal importance sampling 2026-10-25
Let be an unnormalized density, with , and . Given independent draws and an evaluable normalized density supported where , reciprocal importance sampling estimates by . Its expectation is and its variance is . Broad tails of relative to can make this infinite; reversing the importance sampling ratio reverses the dangerous tail mismatch. Taking the reciprocal of the estimate is generally biased for .
Reciprocal moments of an estimated variance 2026-11-05
Reciprocal moments of an estimated variance determine whether a precision estimate such as has a finite expected value or variance. Small positive values of a sample variance become large values after inversion. An unbiased estimator of a variance therefore need not give an unbiased or even integrable estimate of its reciprocal.
For , let be independent normal random variables with common mean and variance . For the unbiased sample variance , put . Then has a chi-squared distribution. Its probability density function is for .[ref]
For any real , direct integration and the Gamma function giveThe convergence restriction is essential: near zero the integrand has exponent , so the integral diverges when . Substitution of a negative Gamma function argument into the displayed quotient does not define a nonexistent moment.
Rescaling gives the useful general resultIn particular, the mean reciprocal exists for , while its variance requires :For the mean is , the variance is and the coefficient of variation is . For the mean is finite but the variance is infinite. For even the mean diverges. Increasing the number of groups does not repair a nonexistent moment of each raw reciprocal estimate.
Inverse-variance weighting commonly normalizes these reciprocals. Those normalized weights lie between zero and one, so their moments can exist even when the raw reciprocal moments fail. This distinction matters in uncertainty in estimated inverse-variance weights: neither the moment calculation nor a plug-in precision formula alone establishes the accuracy of the final weighted estimator. The results here use the exact normal-sample law; a nonnormal sample requires a different analysis.
Reference identity in algebraic uncertainty cancellation asks whether two appearances of a reference are the same measured quantity or merely readings given the same descriptive name. If a transfer is but a correction subtracts , thenThe last difference vanishes identically only when the same reference value is reused. Separate acquisitions may be correlated, independent, or affected by drift; sharing a label does not decide their covariance.
For independent and , also independent of the reference pair, propagation givesWith equal reference standard uncertainties and correlation coefficient , the residual reference contribution is . It is zero for identical reused values, for independent readings, and is not generally zero just because the displayed numbers agree. The GUM requires significant correlations from common inputs to be included.[ref]
An original acquisition model separates what cancels:Here is a common uncertain calibration contribution, while are independent repeat errors. Their difference removes both and , but leaves . If mg and each repeat error has standard deviation 2 mg, each reading has variance and their covariance is . The difference therefore has variance , not zero.
Adding independent mg and mg gives . Reusing exactly one recorded reference value in both places instead gives . These constructed cases have different data-reduction histories despite similar-looking equations.
Reference lineage in covariance models records acquisition identifiers, shared calibration inputs and coefficients in each derived quantity. A calibration change or time-dependent drift must appear in that model if significant. Algebraic simplification is valuable only when it preserves the identity of the underlying inputs; renaming two acquisitions with the same symbol does not make their errors cancel.
Reference material homogeneity 2026-11-05
Reference material homogeneity concerns whether variation of specified properties within or between usable portions is small enough for the intended measurement purpose. Homogeneity is property-specific and depends on the portion taken: uniform colour need not establish uniform refractive index, composition or another measured property. Sufficient homogeneity is part of the definition of a reference material.[ref]
International vocabulary of metrology by JCGM. VIM3entry5.13, definition
An original study takes several bottles from a batch and measures multiple portions from each, with the measurement order varied to avoid confusing bottle order with instrumental drift. A simple measurement model iswhere describes bottle-to-bottle variation and includes variation among sampled portions and measurement noise. For independent zero-mean contributions, the variance of a single observation isRepeatedly reading one portion mainly probes repeat measurement scatter. It does not replace sampling different portions or bottles. Separating material variation from measurement repeatability requires a sampling and replication design appropriate to both.
Suppose an original fitted model gives a between-bottle standard deviation in refractive index and within-bottle measurement scatter . Their combined standard deviation for one observation is under the independence assumption. Averaging more readings from the same bottle can reduce the second contribution, but does not average away that bottle's common contribution. These are illustrative model values, not certificate data.
NIST's archived SRM1922 characterization tested multiple oil portions using minimum-deviation measurements and an Abbé-type refractometer.[ref]
Certificate for Standard Reference Material 1922 by National Institute of Standards and Technology. 2009certificate, p1, Certified Values and Uncertainties, homogeneity tests on multiple aliquots
Sampling different portions answers the homogeneity question directly; reference material stability separately addresses changes over time. A report should retain the sampled portion size and conditions so that a user can judge whether the evidence applies to their own use.Reference-versus-sample allocation in repeated isotope-ratio measurements distributes a finite measurement budget between sample readings and the reference used to calibrate them. Suppose independent sample and reference readings have first-order gain-error variances and . Averaging sample readings and reference readings gives log stiffness-ratio variancewhere is the variance of a differential calibration error shared by all ratio estimates. The shared-reference floor in averaged isotope stiffness ratios explains why that last component does not average down. Covariance propagation must reflect which observations and calibration influences are actually shared.[ref]
Let one sample reading cost budget units and one reference reading cost , with . Treating counts continuously for the initial optimization, a constraint multiplier givesThe counts are proportional to standard uncertainty divided by the square root of cost, rather than simply to each variance. A budget-independent common component changes the attainable floor but not this optimum ratio under the stated model.
For an original example, take equal costs, , and 40 total readings. The optimum is 30 sample and ten reference readings. Without a common component, the predicted log-ratio standard uncertainty is about , compared with for twenty of each. A common standard uncertainty raises those values to about and , respectively.
An integer allocation near a continuous isotope-reference optimum must compare feasible neighboring count choices and any acquisition-order constraints. More reference readings help only to the extent that their errors are independent; repeatedly measuring an unchanged unknown calibration bias does not create independent references. If reference drift depends on elapsed time, a single constant cost and common-variance term may be inadequate.
In practice, record which reference observations calibrate which sample spectra and account for any correlated peak fitting, isotope-composition uncertainty and spectral offsets. The variance and cost model above is a chosen allocation problem, not an experimentally established precision specification for an isotope measurement system.
If a set partition refines a rectangle of a graph into rectangles , write for its edge density of a bipartite graph. The weighted variance identity isIt follows by expanding the square and using . In particular, equitable regularity energy cannot decrease under refinement. If a witness rectangle has relative sizes at least and edge density of a bipartite graph discrepancy greater than , its contribution, after division by the squared total vertex count, exceeds .
For air approximated by index one, , so . Taking the variance gives the displayed first-order error formula, with angles and their errors in radians. Independence removes the covariance term; hard error bounds require a different combination rule. Common angular-zero errors can produce covariance and should not automatically be added in quadrature.
Regression coefficient test power ignores nuisance coefficients Created 2026-10-05 Updated 2026-11-05
In a fixed full-rank normal linear model, the studentized estimator of one coefficient has a noncentral t-distribution with noncentrality equal to that coefficient divided by its standard deviation. The fitted and residual orthogonal projections of the Gaussian noise are independent. Therefore the statistical power of its exact Student's t-test does not depend on the other regression coefficients, with design and variance fixed.
Regression diagnostics Created 2026-10-06 Updated 2026-10-07
Regression diagnostics check assumptions and influential observations in a fitted regression function. Residual-versus-fitted plots can reveal omitted mean structure or nonconstant variance; a quantile-quantile plot checks a specified error distribution. Regression leverage measures unusual predictor configurations, and Cook's distance combines leverage and residual size to measure coefficient sensitivity. Independence may require checking collection order and the study design in addition to residual plots.
Regression error 2026-11-05
In a linear regression model , the regression error is the unobserved discrepancy between the response and the model's systematic component. The assumption makes the conditional expectation of given ; it does not follow merely from writing the equation.
For a scalar response , an observation has error one if the stated coefficients are the true model coefficients. A regression residual instead uses fitted coefficients: . Errors and residuals therefore answer different questions. In ordinary least squares, holds algebraically when the fit exists, even though need not be zero for a realized statistical sample. Assuming independence or constant variance of the errors is a further modeling choice.
Regression outlier 2026-10-07
A regression outlier has a response unusually far from its fitted conditional mean. A standardized regression residual adjusts for regression leverage; an externally studentized residual also estimates noise variance with the observation deleted. Selecting the largest residual among many observations changes its reference distribution. An outlier need not be an influential observation, and unusual valid data should not be discarded solely for being unusual.
Regularization Created 2026-09-24 Updated 2026-11-05
Regularization constrains or penalizes fitting to control the complexity or instability of a learned rule. A penalty formulation iswhere is empirical loss function and measures a chosen notion of complexity. Ridge regression penalizes squared coefficient size, the Lasso penalizes the L1 norm, and a roughness penalty penalizes derivatives of a fitted function. Early stopping can regularize an iterative fitting procedure without an explicit added penalty.
For a scalar observation with centered noise of variance , minimizing gives . Its mean squared error is : shrinkage reduces variance while introducing bias. This illustrates why regularization can improve prediction, but does not improve risk at every parameter value for every tuning choice. Cross-validation or a validation set can choose strength while a reserved test set assesses the fitted procedure.
The lowest even quantum harmonic oscillator quantum state has density , and the lowest odd quantum state has density . Since , integration by parts gives whenever the boundary terms vanish. For , this proves the displayed identity. Odd moments vanish in both quantum states by parity; even moments quantify the larger spatial spread in the odd quantum state. In particular the relative-coordinate variance is three times as large, and the odd density vanishes at particle coincidence.
