An affine pressure-tap coordinate calibration converts an instrument coordinate to a physical coordinate through an affine map, . The offset fixes the datum and the nonzero scale fixes the length conversion. Its adequacy is a question about the actual viewing or measuring geometry, rather than a consequence of having fitted two points.
For two distinct instrument readings at known physical references ,
An original imaging example takes the leading edge at pixels and , and the trailing edge at pixels and m. Hence m/pixel and m. A tap at pixels is at m, or dimensionless chord position .
If all three features share the same affine map, the dimensionless coordinate is simply
Changing the displayed coordinate to gives readings 230, 1030 and 470 pixels and leaves . This is the exact pressure-tap datum cancellation; it does not assert that the original feature readings are known without uncertainty.
For example, with , the sensitivity coefficients of with respect to the tap, leading-edge and trailing-edge readings are
Their sum is zero, as required for invariance to a common offset. At the example readings they are per pixel. Use their full covariance matrix when propagating uncertainty, because independently attaching three origin errors would destroy the cancellation.[ref]
An affine pressure-tap fiducial check uses an additional independently located mark. At pixels the fitted map predicts m. A measured location m leaves a 2 mm discrepancy to investigate. Two fitted reference points necessarily have zero regression residual; that fact cannot validate the map between them. Perspective, surface curvature, optical distortion and reference-location uncertainty can all matter. A proposed calibration records extra marks across the image and compares residual patterns before using the map for surface pressure integration for airfoil lift.
Arrhenius equation 2026-10-13
The Arrhenius equation models the temperature dependence of a rate constant by
Here is the pre-exponential factor, the activation energy, the molar gas constant, and the absolute thermodynamic temperature. In its original form, and are held constant over the range being modeled. It is an empirical relationship, not a universal rule for every reaction mechanism. See the Arrhenius equation.
Eliminating between two positive temperatures gives the useful prediction
For , warming increases . The logarithmic sensitivity is , so there is no universal doubling per ten degrees. The same rise from to gives a factor for , versus for .
Figure 1. . Both models are anchored at , so they have different pre-exponential factors. Left: the higher activation energy produces a larger relative change with temperature. Right: the natural logarithm of the dimensionless ratio is linear in inverse temperature. These exact model curves are not experimental data. Details.
The useful simplification is that a strong change in reaction rate becomes a straight-line relation after changing variables. Measurements at several temperatures can estimate the slope and test whether one pair of constants describes the chosen range. Curvature can indicate a failure of that approximation; fitting a line does not uniquely establish a microscopic reaction mechanism.
Svante Arrhenius's 1889 paper on sucrose conversion through catalysis by a chemical acid compared the strong observed temperature dependence with the much smaller increase in collision frequency. He discussed van't Hoff's earlier temperature relations and proposed a small population of hypothetical active sucrose, replenished from ordinary sucrose, to interpret the dependence. The inspected translated excerpt derives an inverse-temperature relation in its equations (6)–(7) and then applies it to the reaction rate. That historical hypothesis is distinct from a modern transition state description. See the On the Reaction Velocity of the Inversion of Cane Sugar by Acids. The excerpt omits parts of the original paper; it does not provide a complete original apparatus description or dataset.
The pre-exponential factor must have the same units as the rate constant, since the exponential function supplies a dimensionless quantity. For a rate law , both and have units . Thus a first-order reaction uses , whereas a second-order reaction uses . Calling a frequency factor does not give it frequency units for every order of reaction. Its value includes the physical and geometrical factors allowed by the chosen reaction mechanism; equal activation energies need not mean equal reaction rates.
Figure 2. . Both chosen first-order reaction models have and the same exponent parameter . Blue uses a constant pre-exponential factor; orange multiplies it by . All markers are synthetic evaluations, not observations. Left: natural logarithm of against . Right: the corresponding apparent activation energies, and in . The dashed orange tangent at has numerical slope on the displayed horizontal scale. Details.
To read the blue Arrhenius plot, define the dimensionless quantity and . Its numerical slope is , giving . The slope against , by contrast, is . The orange tangent at gives ; its finite-interval slope between and gives about . The distinction between a local derivative and a fitted interval matters even with perfectly noise-free synthetic evaluations.
Two measured positive rate constants give a two-point estimate
Several temperatures let linear regression test the constant-parameter approximation: inspect the regression residuals, keep the same rate law and composition convention, and account for uncertainty in both and . A straight line over a short range supports a useful local approximation; it does not uniquely identify a reaction mechanism. A fixed change of units or of shifts the intercept without changing the inferred activation energy.[ref]
Augmented inverse-probability-weighted estimator Created 2026-09-24 Updated 2026-11-05
An augmented inverse-probability-weighted estimator averages a regression prediction plus an inverse-probability-weighted residual. In the treatment setting of doubly robust estimation, let estimate the conditional outcome mean and the probability of receiving treatment . Then
The first term predicts outcomes for everyone; the weighted residual corrects prediction errors using observed outcomes from the corresponding treatment group. The conditional-expectation calculation in doubly robust estimation makes the remaining bias a product of outcome-regression and propensity-score errors. An exactly correct outcome regression makes the residual correction mean zero even with an incorrect propensity score; a correct propensity score makes that correction repair the regression bias.
The nuisance estimates are commonly obtained by cross-fitting, with each observation evaluated using functions trained on other observations. Both treatment probabilities must remain positive on the population being studied, and small fitted probabilities can make the variance very large. Double robustness concerns consistency under one correct nuisance model; it does not automatically deliver valid finite-sample or asymptotic confidence intervals.
Calibration residual 2026-11-05
A calibration residual is the difference between a measured response and the response predicted by a fitted analytical calibration function at the corresponding reference input. With observed signals , reference concentrations and fitted function , choose the convention
A calibration residual is an observable diagnostic of a fit; it is not the unobserved measurement error relative to a true response. Plotting regression residuals against reference input, fitted response and measurement order can reveal different model failures.[ref]
For an original example in chosen consistent concentration and signal units, take reference inputs and responses . Ordinary least squares gives
The calibration residuals sum to zero, as expected for an ordinary least squares fit with an intercept, but they are not individually zero. The sum of their squares is ; with three observations and two fitted coefficients, the residual estimate of noise standard deviation is signal units. One residual degree of freedom provides a weak estimate, rather than strong evidence for the assumed noise law.
A curved residual pattern suggests a missing response term, while changing scatter can suggest nonconstant noise variance. These diagnoses interact: an incorrect response curve itself can create apparent changes in scatter. Replicate standards and measurements across the intended range help distinguish them.[ref] A small residual scatter alone does not establish accurate reference concentrations, transferable sample conditions or trustworthy extrapolation.
Coefficient of determination Created 2026-10-06 Updated 2026-10-14
In an ordinary least squares regression with an intercept and a nonconstant response, . It measures the fitted proportion of sample variation relative to an intercept-only baseline. It increases when regressors are added, even if those regressors are unhelpful for prediction. A high value does not establish the assumptions of a normal linear model or rule out a systematic regression residual pattern.
For the five-point example in linear regression, the total centered sum of squares is and the residual sum of squares is , so . This summarizes the fitted reduction in squared error. It does not record the order or shape of the regression residuals, which are worth examining in a residual-versus-fitted plot.
Here is a fully constructed illustration. Use and the same fitted line for both datasets. Add respectively the residual vectors
Both sum to zero, both have zero inner product with , and both have squared norm . Their ordinary least squares fit is therefore exactly the same line, with the same . But the first residual is , a systematic quadratic shape, while the second changes direction across the predictor values. These examples show directly which information the scalar discards; neither constructed pattern proves a random-error model.
Figure 1.
Two constructed datasets share a fitted line and R squared but have different residual shapes
. Details.
Conditional multivariate normal distribution Created 2026-09-24 Updated 2026-11-05
Partition a jointly multivariate normal distribution into vectors and with means and covariance blocks . If is invertible, then the Conditional multivariate normal distribution is
The covariance is a Schur complement. To see why, subtract the linear prediction from : . Direct covariance calculation gives . Since and are jointly Gaussian, they are independent; the covariance of is exactly the displayed Schur complement. Conditioning therefore changes the mean but leaves that residual covariance fixed.
For two standard normal variables with correlation , has mean and variance . If , the conditional law is degenerate: the same formula still gives a zero variance. A singular requires conditioning on its supported subspace rather than taking an ordinary matrix inverse.
A concrete laboratory route to an Arrhenius plot is to infer a rate constant at each controlled temperature from a changing electrical conductivity. The Technical University of Munich's Activation Energy of a First-Order Reaction supplies a teaching protocol, rather than a published set of measured rate constants.
The manual follows tertiary amylbromide conversion, which produces conducting ions. A measuring cell connects to a thermostat; an immersed electrical conductivity probe also measures solution temperature, and a magnetic stirrer mixes it. Use of by volume ethanol in water, equilibrate, then add of tertiary amylbromide. Record the baseline before addition and set time zero at addition. Repeat with fresh mixtures at , , and , each within ; record for minutes at the first two temperatures and until completion at the last. Clean the cell between runs and keep the added volume fixed. The completed mixture supplies endpoint electrical conductivity measurements at the other temperatures: one endpoint reading cannot be used unchanged at every temperature.[ref]
To see what is inferred from the readout, let , , and be the baseline, current, and completed-mixture electrical conductivities at the same temperature. Assume their baseline-subtracted changes are proportional to converted concentration. Combining that measurement model with a first-order reaction gives
Thus the electrical conductivity increases toward an endpoint, while the remaining reactant fraction decreases. Fit against time to obtain ; then fit against in kelvin to estimate an apparent activation energy. This baseline-aware derivation states the measurement assumptions explicitly; it does not infer a rate constant from the steepness of a raw conductivity trace alone.
The controls distinguish a faster chemical reaction from a larger instrumental response. Measure the actual solution temperature, retain the same solvent composition and initial reactant concentration, and compare endpoint and baseline at that temperature. As further diagnostic checks, repeat runs and inspect regression residuals for departure from a first-order reaction. These checks follow from the model: electrical conductivity also depends on how ions move, a baseline drift can mimic conversion, and a mixing delay can distort early points. Near completion, a small endpoint error produces a large relative error in and its natural logarithm. A successful fit supports an apparent rate model for this solvent and interval; it does not directly image a transition state or identify a unique reaction mechanism. No measured or is supplied here.
Equilibrium residual-life distribution Created 2026-09-24 Updated 2026-11-05
The equilibrium residual-life distribution describes the remaining wait to a renewal when observing a stationary renewal process at an ordinary time. For positive interarrivals with finite mean , it has survival function and density
The density integrates to one by the Tail-sum formula for expectation. It exists even when the original lifetime law has atoms. A deterministic lifetime gives a uniform residual on ; an exponential lifetime gives the same exponential distribution as the original lifetime.
For the discrete chain convention in residual lifetime Markov chain, the invariant probabilities are instead for . A deterministic integer lifetime gives a uniform law on . These formulas describe observation at equilibrium; they do not make the first lifetime of an ordinary renewal process already have that residual law.
Extra sum of squares Created 2026-10-06 Updated 2026-11-05
For nested normal linear models, the decrease in residual sum of squares obtained by adding coefficients is their Extra sum of squares. Dividing by the number of added coefficients and by the full-model residual mean square gives the partial F-test. A sequential analysis of variance attributes this decrease according to the specified order of entry; it need not equal the partial comparison when other terms are retained.
For a full-rank ordinary least squares design matrix, the hat matrix and residual projection satisfy . Hence fitted values and regression residuals are perpendicular as observed vectors. Under isotropic errors their cross-covariance matrix is zero, and under a normal linear model the two vectors are independent.
Fitted values Created 2026-09-24 Updated 2026-10-14
Fitted values are the predictions at the predictor values used to fit a regression model. In linear regression with an explicit regression intercept,
Here is the slope design matrix, and is the regression residual vector. The fitted values estimate the regression function at the observed predictors; they need not equal the observed responses, which may also contain random errors. Prediction at a new predictor vector uses with the same centering and scaling used in fitting. A prediction interval addresses uncertainty in a new response rather than just displaying this point prediction.
For the five constructed observations in linear regression, ordinary least squares gives fitted values . The ridge regression fit at instead gives . Both fitted vectors have sample mean because the regression intercept is unpenalized; their regression residuals differ and their residual sums of squares are and , respectively.
For ordinary least squares, the fitted values are the orthogonal projection of onto the column space of the complete design matrix, including its regression intercept column. They remain unique even in rank-deficient ordinary least squares, although the regression coefficients can be nonunique. The regression residuals are then orthogonal to every column.[ref] With centered slopes, ridge regression instead satisfies , so the regression residuals need not be orthogonal to the slope columns. Their sum is still zero because the regression intercept remains free.
For fixed predictors and scales and a fixed positive penalty, ridge regression gives a linear smoother,
For with at least one nonzero slope direction, this matrix is not an orthogonal projection matrix: applying it twice shrinks that direction again. Its effective degrees of freedom are , where are the positive singular values of ; the extra counts the free regression intercept. If the penalty is selected using , the resulting fitting rule need not be a linear smoother, since its chosen also depends on .
Gaussian error Created 2026-10-31 Updated 2026-11-05
A Gaussian error is a random error term with a Normal distribution. The exact least-squares conclusions require a specified model: for fixed-design linear regression, let
where the design matrix has full column rank. The error vector thus has independent zero-mean Gaussian coordinates of common variance. For fixed , maximizing the likelihood over is equivalent to minimizing .
The least-squares estimator and its exact sampling law are
This follows because a linear transformation of a Gaussian vector is Gaussian. It is an exact finite-sample law, rather than merely Asymptotic normality.
Zero mean and common variance alone do not imply Gaussian errors or exact normal coefficient estimates. Even with Gaussian errors, fitted residuals are not independent copies of them: with , the residual vector is and has covariance , usually with correlated coordinates and reduced rank. A Normality assumption must identify whether it concerns errors, responses or another latent variable.
Choose the five numbers as an original illustration. Their sample median is 2 and their sample mean is 4. The sum of absolute losses at the sample median is
If these numbers were modeled as an independent and identically distributed Laplace distribution sample, the joint maximum-likelihood estimates would therefore be and . Using the sample mean instead gives , so it does not maximize this likelihood function.
Now replace only 10 by 100. The sample median stays at 2, while the sample mean becomes . The minimized sum of absolute losses rises to , and the fitted scale rises to . Stability of the fitted location therefore does not mean stability of every fitted parameter.
Figure 1.
Chosen five-value example: moving the largest value leaves the median fixed but moves the mean
. Details.
The left panel subtracts each objective's minimum so the two absolute loss curves can be compared despite the different totals. The right panel tracks the sample mean and sample median as the largest chosen value changes. This is an outlier sensitivity illustration, not actual observations or a check that a Laplace distribution fits any population. In applications, regression residuals and the data-generating assumptions need separate examination; a stable sample median alone does not validate a statistical model.
Laplace regression Created 2026-09-24 Updated 2026-10-18
For linear regression with fixed predictor vectors , assume conditionally independent responses with a common scale :
Each conditional median and mean is . The log-likelihood function is , so maximum likelihood estimation minimizes the sum of absolute regression residuals:
This is least absolute deviations. With only an intercept, it reduces to the Laplace location and scale maximum likelihood problem, so a sample median is a fitted intercept. Provided the minimum , the scale estimate is ; the conditional standard deviation is . If the scale is denoted instead of , the conditional standard deviation is .
Minimizing a sum of absolute losses does not automatically give a unique coefficient vector. Even an intercept-only fit to an even sample size can have an interval of solutions, and redundant predictor columns can introduce further nonuniqueness. If a fit makes every regression residual zero, the joint likelihood function is unbounded as , so there is no finite joint maximum-likelihood estimate with positive scale.
The linear growth of absolute loss limits sensitivity to an extreme response at fixed predictors compared with squared-error loss. It does not protect against every outlier: an extreme predictor can have large regression leverage. The numerical example above changes a response in an intercept-only problem, and makes no general claim of immunity to such predictor changes.
Linear least-squares projection Created 2026-09-24 Updated 2026-11-05
A linear least-squares projection is the affine predictor that minimizes mean squared error. For real random variables with finite second moments and , minimizing gives
The residual satisfies and . Consequently it is orthogonal, in the inner product , to the span of the constant variable and . For any other affine predictor , expanding the square yields
because the cross term vanishes. This proves optimality and uniqueness of the predicted random variable. The residual variance is the Schur complement covariance. No normal distribution is needed. For several predictors, the same orthogonality argument gives the normal equations for Linear least squares; a singular predictor covariance can make coefficients nonunique while leaving the projected prediction unique.
Linear regression Created 2026-09-24 Updated 2026-10-14
Linear regression models a response's conditional expectation as a linear combination of specified predictors. In matrix form,
where the design matrix describes the observations, contains the regression coefficients, and contains the errors. The fitted values estimate the response means. The model is linear in its regression coefficients; its columns may include powers, interactions, or other transformed predictors, so polynomial regression can fit curved relationships using the same machinery.
For a concrete example, take five constructed observations
An ordinary least squares fit with an intercept gives the line . Its regression residuals are and its residual sum of squares is . A one-unit increase in the predictor corresponds to a one-unit increase in the fitted response. This is a description of this fitted association; interpreting a change as a causal effect requires the assumptions of a causal inference problem.
Figure 1.
A five-point least-squares fit separates the fitted trend from residual deviations
. Details.
The geometric reason for the fit is simple. The possible mean vectors form the column space of . Ordinary least squares chooses the orthogonal projection of the data vector onto that vector subspace, so the remaining regression residual is orthogonal to every permitted fitted direction. The least-squares normal equations express this orthogonality as . The data's explained component and unexplained component are therefore separated by the same geometry used in a Hilbert space.
The optimization definition makes sense for any observed data. Quantifying the uncertainty of the regression coefficients needs additional assumptions about the sampling and errors. With fixed of full column rank and mean-zero errors with covariance matrix , ordinary least squares is unbiased and has covariance ; the Gauss-Markov theorem compares it with other linear unbiased estimators. A normal linear model additionally makes Student t-distribution intervals and tests exact in finite samples. Model adequacy and predictive assessment remain relevant even when these calculations are available.
Two historical settings help explain the subject's appeal. Adrien-Marie Legendre's Nouvelles méthodes pour la détermination des orbites des comètes (1805) includes an appendix on the method of least squares: observations give more equations than unknowns, and squared errors provide a criterion for reconciling them. Its astronomical setting shows the motivation for combining imperfect measurements. [ref] Francis Galton's Regression Towards Mediocrity in Hereditary Stature (1886) examines the association between parental and offspring heights and the tendency of group averages toward the population average. This is a historical use of the regression name; the modern linear model is a general statistical formulation. [ref]
Log-linear exponential-decay fit Created 2026-10-18 Updated 2026-10-25
An exponential decay rate becomes a straight line in and , where is a fixed reference rate. Constant small fractional rate errors become approximately constant vertical errors by propagation of uncertainty. With independent Gaussian log-rate errors, linear least squares gives and .
The covariance matrix is obtained by applying the variance rules to these linear estimators:
Transforming back gives , , and . If a candidate lifetime fixes the slope , the intercept must be refitted as . The profiled objective obeys
This follows by separating the centred slope residual from the unrestricted regression residual, whose inner product with the design columns is zero by the normal equations for linear least squares. Known supplied measurement errors are not rescaled by the observed reduced chi-squared. This log-error model is an approximation for proportional errors, not an exact replacement for every raw-count likelihood.
Measurement of psychological resilience Created 2026-10-25 Updated 2026-11-05
Measurement relating documented adversity to subsequent adaptation. A design specifies adversity exposure, outcome, baseline functioning and follow-up; capacity questionnaires, trajectories and adversity exposure-conditioned regression residual scores operationalize different questions.
Minimizing gives the displayed equations by differentiation. Equivalently, the regression residual satisfies : it is orthogonal to every column of the design matrix. If has full column rank, the minimizer is unique and the fitted values are the orthogonal projection of onto its column space. A rank-deficient design matrix can give multiple coefficient vectors, while the projected fitted vector is still unique.
For an intercept and one predictor, put and . The equations give and , provided . In the five-point example in linear regression, , , and , giving and .
The normal equations describe the mathematical optimum, but explicitly forming and inverting can amplify numerical problems. A QR decomposition computes the same least-squares solution through orthogonal coordinates; a singular value decomposition also exposes nearly dependent columns. This connects statistical fitting to numerical linear algebra and the choice of a stable algorithm. [ref]
Normality (statistics) Created 2026-10-20 Updated 2026-11-05
A normality assumption specifies a normal distribution for the relevant observations, errors, or latent effects. These are different assumptions: normal latent random effects do not imply that the observed response is normal. Estimated regression residuals or fitted random effects are not independent copies of the underlying errors or effects.