Relative variance 2026-11-05
For a random variable with finite variance and nonzero mean, its relative variance is . It equals the square of the coefficient of variation under the usual positive-mean convention. Multiplying by a nonzero constant leaves unchanged; adding a constant generally does not.
For an average of independent identically distributed observations with mean and variance , the relative variance is . Chebyshev's inequality gives , explaining its role in self-averaging. At zero mean the ratio is undefined, so another reference scale is needed.
A spatial image after a chosen dark interval is a different readout from continuously recording arrivals at a fixed probe. Here is an original idealised fluorescence imaging model: release a dilute cloud, let it move freely, then apply a short uniform probe pulse. Neglect collisions, residual forces, attenuation and motion during the pulse. Use each image's cloud centre to remove mean drift. Uniform gravity shifts that centre without changing the centred variance.
For each atom let along a horizontal imaging axis. Expanding the square about the expected values gives exactly
No Gaussian distribution is needed for this variance identity. A jointly Gaussian distribution of initial positions and velocities does make the later density Gaussian. For a separable three-dimensional cloud viewed along , integration along the viewing direction gives the column density
Here is atom number. Integrating this image along leaves a one-dimensional Gaussian distribution with standard deviation . This is a projected density measurement, not a three-dimensional radius or a direct velocity histogram. Under the uniform, optically thin probe assumptions, expected fluorescence is proportional to this column density. A known, independent point spread function with finite variance adds its own variance to the measured width; probe duration and intensity gradients need separate treatment.
For a Maxwell-Boltzmann distribution, , where is atomic mass and the Boltzmann constant. For a nonthermal distribution this defines a directional second-moment temperature, which need not describe every feature of the distribution. A covariance between position and velocity can initially narrow the cloud even though is positive. The physical covariance matrix requires .
Choose , and , giving correlation coefficient . With in milliseconds, the numerical variance in square millimetres is . Its minimum occurs at , with standard deviation . These chosen values satisfy the covariance matrix constraint; early contraction is compatible with free motion.
Generate thirteen synthetic variance observations at equally spaced times from to , adding independent Gaussian noise of standard deviation . Linear least squares fits all three coefficients in . The example yields , and . With the chosen approximate sodium mass , the true is and the fitted value is . This is a fit to generated width data, not historical observations.
Figure 1.
Cloud expansion with an initial position–velocity covariance
. Original synthetic variance observations and ideal column-density marginals. All parameters and noise are chosen here. The three-dimensional Gaussian is integrated along the viewing axis and then along the image's other axis. The profiles omit imaging blur. Details.
If the same data are fitted to , forcing , the inferred falls to . The omitted term is partly absorbed into both fitted coefficients. In noiseless linear least squares with a free intercept, its precise effect on the quadratic coefficient is
where the subscript denotes equal-weight moments over the chosen observation times, using one consistent unit system. Since the ratio is positive for distinct nonnegative times, negative biases this estimate downwards. An incorrect release-time origin can also generate a linear term, so fitting it does not by itself identify a physical initial covariance. Timing calibration, spatial resolution, residual magnetic fields and probe-induced recoil energy remain separate controls. Neither a cold-looking image nor a single long expansion time identifies all three coefficients.
Repeat-count targets under shared model error choose how many independent readings are worth averaging when part of the uncertainty is common to every repeat. In the local measurement model , let the shared discrepancy have variance , and the independent repeat noise have variance . Assuming the two components are independent,
The common term produces covariance between readings and survives averaging. The GUM's correlated-input rule retains precisely such shared contributions.[ref]
For a desired total standard uncertainty , solve the inequality before choosing the integer repeat count:
If , a target equal to cannot be attained with any finite repeat count, and a target below is unattainable within this model. This is the feasibility threshold for averaging under shared uncertainty.
In an original body surface area example, take and . A target needs ten repeats; needs one hundred. A target of instead requires reducing the shared component or changing the model. Merely collecting more readings is insufficient.
Two relative criteria illustrate why the target must be explicit. Requiring repeat-noise variance to be at most gives : at , 25 repeats suffice. Requiring total standard uncertainty within 1% of its floor instead gives
which needs 13 repeats with the same stipulated numbers. A 1% variance criterion and a 1% standard-uncertainty criterion are different requirements.
The calculation assumes independent repeat noise and a justified shared-error model. Repeated observations alone cannot estimate an individual fixed discrepancy; independent reference measurements or a suitable validation population are needed to constrain it.
Imagine an instrument measuring the same fixed quantity twice on each unit. Across units the persistent signal is a latent variable; the two measurement errors are fresh, mutually independent random variables, also independent of . Assume they have mean zero and common finite variance , while has mean and variance . Then
Thus the Pearson correlation coefficient is when . Both measurements retain the same marginal mean and variance. Repetition does not make the entire population collapse toward ; it changes what we expect after selecting on .
For , further assuming normal distributions for and both measurement errors gives the conditional expectation
Since , this predicted reading is closer to than is. Independence and zero mean errors alone do not justify this linear conditional expectation: without the normal distribution assumption the same expression is the linear least-squares projection, and the actual conditional expectation can be nonlinear. Correlated errors, a changing signal, or different error variances require their own model; a noisy reading by itself supplies no universal shrinkage factor.
Replication in experimental design Created 2026-10-07 Updated 2026-10-18
Replication in experimental design repeats a treatment on multiple independent experimental units, so that variation between those units can inform a treatment contrast and its standard error. Repeated measurements on one experimental unit can improve measurement precision without providing additional independent treatment replication. The experimental unit is identified by where the treatment is applied, rather than by how many responses are recorded.[ref]
For a constructed example, grow one plant in each of eight separate pots. Apply treatment A to four pots and treatment B to four pots, without sharing drainage water between pots. Measure the length of three leaves per pot. The pot is the experimental unit; the leaf is the observational unit. There are four treatment replicates per group, eight experimental units overall, and measurements. The three leaves in one pot are nested measurements, because they share the same plant and treatment application. The setup assumes that pots do not affect one another; simply putting them in separate containers does not establish statistical independence.
Figure 1.
Constructed two-treatment design: eight pot experimental units contain twenty-four leaf observational units
. Each outlined pot receives one treatment application and supplies three nested measurements. A pot label marks an experimental unit, not a measured response value. Details.
Randomization chooses which experimental units receive which treatments; replication supplies multiple experimental units receiving the same treatment. These solve different problems. In the balanced allocation above, choose the four A pots uniformly among the possible subsets. A particular pot has probability of receiving A. Given that one pot receives A, a second receives A with conditional probability , rather than . Thus the assignments are not independent random variables, even though the experimental units can supply independent responses conditional on those assignments in a suitable statistical model.
An actual ecology experiment illustrates why the distinction matters. In Stuart H. Hurlbert's account, twelve experimental ponds initially received zero, fifty or 450 fish, with five, four and three ponds respectively. Preparation included assessing variation between ponds before introducing fish; allocation considered both location and existing differences between ponds. The zero-fish ponds supplied the scientific control. These were repeated treatment applications to ponds, not twelve samples from one pond.[ref]
The original article followed the southern California ponds for ten months. Its readouts included zooplankton populations and total phytoplankton; it reported decreases in some zooplankton populations and increased total phytoplankton.[ref] Repeated sampling within a pond helps describe its response through time, while ponds supply treatment replication. Unequal group sizes and initial pond variation matter for the statistical model; the treatment counts alone do not establish equal variances or independent responses.
Reset pedestal variance 2026-11-05
Reset pedestal variance is the acquisition-to-acquisition variance of the output offset associated with integrator reset. In a model where the only fluctuating contribution is an opening charge on fixed capacitance ,
Subtracting the mean of the reset charge-injection pedestal removes its constant component, not the random spread. The physical datasheet distinguishes charge-transfer offset from output noise; its charge-transfer limit alone is not a statistical variance specification.[ref]
Finite blank measurements introduce another contribution. Suppose independent pedestals have mean and standard deviation . Use independent zero-input acquisitions to estimate , then subtract that same blank sample mean from each of new signal acquisitions. With a constant signal and no other noise, the corrected mean has
This variance of a pedestal-corrected sample includes the uncertainty of the shared correction. Its second contribution does not shrink when more signal acquisitions reuse the same blank estimate. Covariance-based uncertainty propagation retains such shared terms.[ref]
For original stipulated , and , the corrected mean has standard deviation . An exactly known pedestal mean would give . Repeated zero-input acquisitions at a fixed delay after switching measure the combined pedestal/readout spread; separating its mechanisms requires additional tests. Slow drift or correlations between acquisitions invalidate the independent-sample formula and must enter the measurement model.
Residual estimate of Gaussian noise variance Created 2026-10-06 Updated 2026-10-14
For a correctly specified full-rank fixed-design Gaussian linear regression with , dividing the residual sum of squares by the residual degrees of freedom gives an unbiased estimator of the noise variance. The projection removes the whole deterministic mean, so the numerator's expectation is . A misspecified subset model need not have this property.
For independent errors of common unknown variance in , the ordinary least squares estimate is . Fitting one parameter leaves residual degrees of freedom. Therefore estimates the common error variance, and estimates the slope variance. For errors with a Gaussian distribution, the slope's standardized deviation follows a Student's t-distribution with degrees of freedom. A common calibration bias does not appear as independent residual scatter.
Residual plot 2026-10-25
A residual plot displays regression residuals against fitted responses, predictor values or observation order. Curvature can suggest an incorrect regression function; changing spread can suggest unequal error variances; ordered runs can suggest dependence. A plot is a diagnostic, not a proof of independence. Dependence confined to higher moments, as for adjacent products of independent Gaussian variables, may call for lagged squared regression residuals rather than a basic scatter plot. Fitting also induces dependence among regression residuals even under independent original errors.
Connect one physical resistor across an ideal lossless capacitor , with no applied source and an ideal readout that draws no electric current. Represent the resistor by its noisy Thévenin equivalent. The resulting RC low-pass frequency response gives the stationary capacitor voltage power spectral density
Its mean square follows by the effective noise bandwidth calculation:
This is also the equipartition theorem: the mean capacitor energy is . Raising makes the low-frequency spectrum taller and its width narrower in exactly compensating proportions. It changes the equilibration time , not the final variance.
The integral to infinity belongs to the ideal classical model. For a physical electrical circuit, assume thermal equilibrium, linear components with negligible extra dissipation, and a corner far below the frequencies where parasitic effects, frequency-dependent electrical resistance, or invalidate the model; the omitted high-frequency tail must be negligible. A finite-bandwidth instrument measures only part of this variance. Its input capacitance contributes to the total , and its finite input electrical resistance introduces both loading and its own thermal electrical noise.
For and , the predicted equilibrium variance is and the root mean square voltage is . At , and . At , and ; the equilibrium root mean square voltage is identical.
Figure 1.
Changing electrical resistance changes noise timescales but preserves
. Original calculated comparison at and . Left: the physical parallel resistor and capacitor and their noisy Thévenin equivalent. Middle: equilibrium one-sided voltage spectra for and have the same full integral. Right: mean square voltage grows toward the same equilibrium value from an exactly discharged initial capacitor. Curves describe ensemble expected values, not measured or simulated voltage traces. Details.
To obtain the growth curve, prepare exactly and connect the resistor at while maintaining its temperature. The RC low-pass filter's causal impulse response is for . Classical ideal white noise with two-sided power spectral density gives
The factor here is the two-sided time-domain noise intensity; the equilibrium spectral integral above used the one-sided . They give the same long-time limit. The variance approaches equilibrium with time constant , although a deterministic initial voltage decays with time constant . If the capacitor already starts in thermal equilibrium, its initial variance supplies the missing exponentially decaying term and the total variance stays constant.
Resolution-limit conditioning in finite-input detector calibration is the sensitivity of a small inferred detector width to uncertainty in two nearly equal measured widths. In finite-input correction in pulse-detector calibration, the nonnegative normalized convolution model gives
The signed relative sensitivities are
They diverge as the measured output width approaches the known input width . Subtracting two well-measured squared widths can therefore be a poorly conditioned inference.
For an original example, and give and . The two relative sensitivity coefficients are about and . A small independent fractional width error can produce a much larger fractional response-width error; a common multiplicative timebase error instead partly cancels because the coefficients sum to one. Correlations must therefore be retained.[ref]
Near the boundary, report uncertainty in the variance difference before transforming it. For a chosen interval , imposing the physical constraint gives a detector-width interval , approximately . A coverage-preserving detector-width bound from a variance interval uses this monotone square-root transformation on the physical intersection. It need not resemble a symmetric error bar around a point estimate. An interval wholly below zero contains no admissible width under this model; replacing it silently with a zero-width result hides the inconsistency.
NIST characterized its detector/cable/oscilloscope chain with an optical pulse much shorter than the receiver response.[ref] That design avoids subtracting two nearly equal broad widths. The variance identity still requires nonnegative profiles and finite second moments; arbitrary full widths at half maximum and responses with negative ringing need a different model.
A response-normalized tolerance for baseline-feasibility residuals measures a pulse estimator's constraint errors against physically chosen response limits. For residual , assign positive tolerances in the corresponding response units and calculate
The criterion means every specified response limit is satisfied. Numerical feasibility residual for baseline constraints explains why a raw unscaled norm alone has no universal physical interpretation.
These tolerances can enforce a deterministic bias budget. If signal and background coefficients are bounded by and , then
The second inequality follows when the normalized test passes. The allocation of a deterministic bias budget among baseline constraints therefore connects a numerical acceptance criterion to a stated physical error allowance.
For an original example, bound the signal amplitude by , constant background by and drift rate by . Choose signal-response tolerance , constant-response tolerance and linear-time-response tolerance . Their maximum bias contributions are respectively , and , totaling . If all actual residuals are half these tolerances, and the corresponding bias bound is .
The normalized test can be invariant to constraint units. Multiplying one equation and its required response by a nonzero factor multiplies its residual by ; replacing its tolerance by keeps the ratio unchanged. The associated coefficient bound changes inversely, so the physical product is unchanged as well.
NIST's before/after pulse baselines supply a measurement setting where residual offset and drift responses matter.[ref] The example bounds and tolerances are stipulated, not specifications of that apparatus. This deterministic leakage test is separate from noise variance, which needs the actual measurement covariance.[ref] Backgrounds outside the assumed family also need separate treatment.
Restricted diffusion 2026-11-05
Restricted diffusion is diffusion whose displacement statistics are constrained by boundaries that particles cannot cross. A reflecting boundary condition models an impermeable wall: particles continue moving, but there is no normal diffusive flux through it. Finite permeability would require a different boundary model.
A physical demonstration uses a liquid confined between solid lamellae. In his first-person account, John E. Tanner described constructing 99 mica layers separated by aluminium-foil strips, removing wrinkles and burrs to obtain regular spacing; Edward O. Stejskal inspected the stack with a stereomicroscope. Pulsed magnetic-field gradients encode displacements in the nuclear magnetic resonance signal, allowing motion normal to the barriers to be compared with less confined directions.[ref] This is a material confinement experiment, not a change in the liquid's chemistry.
For an original one-dimensional model, let particles diffuse in with constant diffusion coefficient :
The wall conditions are homogeneous Neumann boundary conditions for the diffusion equation. Start in equilibrium, so has the uniform distribution on . Its variance is
At long time separation, the positions decorrelate. Consequently the mean squared displacement approaches
At short times relative to , most displacements have not encountered a wall, and the leading scale is the free-diffusion result . For a hypothetical gap , the long-time plateau is and its root is about .
Figure 1.
Equilibrium restricted diffusion saturates in a reflecting interval
. Original Neumann-eigenmode calculation with uniform initial positions; time is and mean squared displacement is divided by . Perfectly reflecting walls give a plateau , while free diffusion continues as . These are model curves, not measurements of the mica stack. Details.
The initial condition matters: fixing instead gives the long-time value . Likewise, echo attenuation in a confined sample must not automatically be converted into a free-diffusion coefficient. Confinement geometry, timing and the selected boundary model all affect the inference.
Retained compound Poisson aggregate Created 2026-10-07 Updated 2026-11-05
Applying a measurable per-claim retention to independent claim sizes preserves the compound Poisson distribution form. The transformed severity is the pushforward measure of the original severity law. With count parameter , the retained aggregate has expected value and variance . Zero retained marks may be removed by Poisson thinning.
If has exponential distribution with expected value , the tail integral formula for moments gives and , where . At matching retained expected value, the excess variance under quota share reinsurance is .
Free bending has two zero-energy filament modes, so it does not define an unrestricted normalizable thermal height distribution. Fix the mean displacement and linear tilt, or project out that affine kernel. On the positive complement, the equipartition theorem gives independent Gaussian mode variances for normalized modes. The resulting covariance is the bending operator pseudoinverse multiplied by . It depends on how the rigid motion is fixed.
Roberto Bruno 2026-11-05
Roberto Bruno is a physicist who studies the solar wind as a natural laboratory for turbulence. INAF's 2018 institutional interview documents his work on nonlinear processes in space plasmas and the American Geophysical Union's recognition of that work.[ref]
His 2024 first-authored study examined proton and alpha particle populations measured by Solar Orbiter. Instead of characterizing all particles by one bulk value, the team used clustering to separate core and beam populations within the velocity distribution. That distinction changed the interpretation of temperature and its directional dependence, because two populations moving relative to one another can appear broader when combined.[ref]
An illustrative calculation shows the issue. Two equally populated, individually narrow groups moving at and have combined mean , but their relative drift contributes to the combined variance. Treating that whole width as random thermal motion would hide the organized relative motion. Measurements of a full velocity distribution can therefore reveal physics that one average speed cannot.
Robertson uncertainty principle Created 2026-09-24 Updated 2026-10-20
The Robertson uncertainty principle bounds the product of the standard deviations of two observables in the same quantum state. For a normalized pure state and self-adjoint operators , write their expectation values and variances as
and similarly for , where is the identity operator. The quantum uncertainties then obey
Here is the commutator, and every expectation value is taken in the same . These standard deviations describe the outcome spreads when and are measured on separate ensembles prepared in that same quantum state. They are not instrument errors or the change caused by a preceding measurement in quantum mechanics. The definition and bound are developed in Barton Zwiebach, Uncertainty Principle and Compatible Observables, 2013, §§1–2.
For unbounded self-adjoint operators, the operator domains matter. Assume , so that both second moments are finite; here means the second moment of the measured probability distribution, equal to , even if is undefined. To use the displayed operator commutator, also assume . Without these product-domain assumptions, the always meaningful version on uses as its right-hand side. These qualifications prevent a formal commutator identity from being applied to an inadmissible quantum state.
The brief proof is Cauchy-Schwarz inequality applied to and :
The final equality uses self-adjointness and the product-domain assumptions. Thus a geometric statement about inner products becomes a constraint on experimental spreads. For nonzero , equality requires to be proportional with a purely imaginary proportionality factor; the equality case of the Heisenberg uncertainty relation turns this condition into a Gaussian wave packet.
The bound is useful because computing a single commutator and its expectation value can constrain two entire outcome distributions. Its right-hand side depends on the quantum state: the spin uncertainty state dependence example gives both a positive saturated bound and a zero bound for the same noncommuting observables. Robertson's paper, titled "The Uncertainty Principle", appeared on 1 July 1929 in Physical Review, volume 34, starting at page 163. [ref].
Robust estimator 2026-11-05
A robust estimator is an estimator whose behavior remains acceptably stable under specified departures from an assumed statistical model. “Stable” must identify a criterion, such as limited bias under contamination, bounded response to a single extreme observation, or resistance to replacing a fraction of the sample. It does not mean that the estimator is accurate under every possible probability distribution.
For example, consider data with . The sample mean is , whereas the sample median, using the midpoint convention for an even-sized sample, is . Sending to infinity makes the sample mean diverge without changing that sample median. The example concerns location estimation: the population median and population mean need not coincide outside a symmetric reference model.
One formal model is Huber contamination:
where is a reference probability distribution, is an unspecified contaminating probability distribution, and is its proportion. A robustness claim can bound an estimator’s worst bias over allowed , rather than only calculating its variance when the reference model is exactly true. Peter J. Huber’s original location paper treats this tradeoff systematically.[ref]
A Huber location estimator illustrates how stability enters the calculation. With a specified positive scale , solve
Large residuals have bounded scores, so one increasingly extreme point cannot contribute an arbitrarily large term to this equation. By contrast, the squared-error score grows linearly with the residual and yields the sample mean. A practical analysis must also state how is obtained: a scale estimate that itself explodes under outliers can undermine the intended protection.
Different criteria answer different questions. An influence function studies an infinitesimal perturbation of a distribution, while a breakdown point concerns a finite amount of sample corruption. A robust estimator can sacrifice some model-specific statistical efficiency to improve performance away from that model, but the amount depends on the chosen tuning and departure. Reporting the target parameter, reference model and robustness criterion makes the claim testable.