Scaled inverse chi-squared distribution Created 2026-09-28 Updated 2026-11-05
For parameters and , the Scaled inverse chi-squared distribution is the law of
The inverse change of variables gives its normalized density
It is an Inverse-gamma distribution with shape and scale . The label is a scale parameter, not the variance of .
Its mean is for , and its variance is for . Below those respective thresholds the positive moments diverge. For example, gives mean three and variance nine.
This is a conjugate prior for the unknown variance of a Normal distribution with known mean . With prior parameters and independent observations, the posterior has and . Multiplying prior and likelihood adds their powers and inverse-variance exponents, which proves the update. An unknown estimated mean requires a separate model and changes the update.
Consider an original balanced-readout model, rather than a fit to the experiment. Let , with a normalized prior probability and . A detected chemical atom has outcome . Define the conditional probabilities
Here is the accumulated quantum phase per photon, the controlled analysis quantum phase, and the fringe contrast. The two probabilities sum to one for every . Actual calibration may also require an offset and asymmetric detection errors; alone does not specify a quantum instrument.
Assume no photon number change during the record and fresh independently prepared atomic meters. For observed outcomes at recorded settings , the Born rule and Bayes' theorem give
The posterior probability is defined only for . Summing its numerator over proves normalization. This is also the update of diagonal elements of a field density operator under a measurement in quantum mechanics; it does not assume that an initial quantum superposition was an unknown pre-existing integer. The observed settings are part of the record, especially when selected adaptively.
For an exact finite example take , , and . An outcome at has likelihood , giving . Another at gives . A subsequent at gives
Thus three atomic records raise the posterior probability of from to , while leaving finite uncertainty. No one of these outcomes identifies with certainty.
Figure 1.
Synthetic atomic records progressively concentrate a photon-number posterior
. Original simulation, not laboratory data: , uniform prior, , , four cyclic analysis phases, and a fixed simulated . Cavity loss is omitted. Individual updates can broaden the distribution even though this record eventually concentrates. Details.
The periodic likelihood function imposes an identifiability limit. If for an integer , then for every setting: the posterior odds between and never change. Varying can resolve a cosine reflection ambiguity within a period, but cannot remove this exact period. Changing can supply additional information. A finite cutoff is therefore a physical assumption to justify, not evidence that excluded high photon numbers were absent. At no outcome changes the prior probability; at neither binary outcome has an exactly zero likelihood.
The no-loss assumption also has a time limit. In an original zero-temperature loss model, let . A Fock state loses its first photon at rate , so its survival probability is and its mean lifetime is for . Several records must be acquired within that lifetime to identify a plateau. When losses matter, first predict using a normalized transition matrix, then apply the atomic Bayesian inference update to . Thermal excitation requires upward transitions too. A delayed or broadened inference after a jump need not indicate that an atom caused the jump.
Signal-subspace preservation for several spectral amplitudes requires a compression to retain every noise-weighted mean direction relevant to those amplitudes. Let have mean , fixed positive-definite noise covariance , and a full-column-rank shape matrix . Compress it as , with full row rank. The original and compressed amplitude-information matrices are
For Gaussian noise these are Fisher information matrices. Without that distributional assumption, they still describe the corresponding known-covariance linear estimation problem.
Whiten the shapes as and put . The orthogonal projection onto the row space of is , giving
This difference is positive semidefinite. It vanishes exactly when every column of lies in that row space, equivalently
Thus preserving one selected shape does not establish preservation of the whole signal subspace. Linear covariance propagation supplies for the compressed observations.[ref]
For an original example, take and shape columns and . Then . Choosing
gives and , hence . Keeping only the first compressed row retains information2 for the first amplitude but destroys sensitivity to the second. An unweighted first row would instead reduce the first information entry to1.8.
At least independent compressed directions are needed to preserve independent shape amplitudes. One construction is , whose compressed signal has mean and covariance . When is invertible, has mean and covariance . In the fixed-covariance Gaussian model, expanding the likelihood function shows that its parameter dependence is retained through .
Unknown nuisance shapes, changing covariance or additional spectral parameters can alter the subspace that needs preservation. The proof concerns the specified family of mean shapes and fixed noise model, rather than guaranteeing that one compression preserves every future spectral question.
For , independent innovations and stationary initial variance , the full likelihood function is
Conditioning on gives a different likelihood.
A sufficient statistic allows the observed likelihood function to be stored through a shorter record, up to a parameter-independent factor. For the Bernoulli sample sum as a sufficient statistic, write
The second likelihood function belongs to the observed count; the first belongs to the observed sequence. Their absolute values differ by the binomial coefficient, but their likelihood ratios between any two interior parameter values agree. A maximum-likelihood estimator can therefore use just .
Figure 1.
Sequence to count: sufficient compression preserves likelihood comparisons
. For , three example sequences share . Recording produces the count likelihood . Normalizing each curve by its maximum makes the two curves coincide. A reverse random draw from the six arrangements restores the full-data probability distribution, using no unknown . Details.
This compression discards which arrangement actually occurred. To reproduce the original experiment's probability distribution, first draw from its binomial distribution and then draw uniformly from the arrangements with that sum. The second step is independent of , so it supplies no additional evidence about . It does not recover the particular discarded sequence. If the scientific question changes to time trends or statistical dependence between observations, the common- independent and identically distributed Bernoulli model may no longer apply, and the ordering can become informative.
Statistical inference Created 2026-09-24 Updated 2026-10-13
Statistical inference uses observed data and a statistical model to learn about unknown features of the process that generated them. In a parametric statistical model, write : the probability distribution depends on an unknown statistical parameter . An estimand specifies the target, such as a mean or a difference of means. A statistic is a function of the observations; an estimator is a statistic used to estimate the target. Uppercase denotes the random data before observation, while lowercase denotes their realized values.
Three questions lead to different outputs. Statistical estimation asks which target value the observations suggest. A confidence interval asks for a range produced by a procedure with calibrated coverage probability. A hypothesis test asks whether the observations depart sufficiently from a specified null hypothesis to enter its rejection region. These outputs complement one another: an estimated effect size describes magnitude, while a standard error describes the sampling distribution of an estimator. A p-value describes a tail probability under the null hypothesis.
The appeal of statistical inference is that it turns variable observations into reproducible uncertainty statements. In the normal mean confidence interval with known variance below, a sample mean of has standard error : both the estimated shift and its precision matter. The calibration depends on the statistical model, including independence and the specified probability distribution. Increasing sample size improves precision under those assumptions; it does not repair a systematically misleading sampling procedure. Bayesian inference provides another framework, using a prior distribution and a likelihood function to obtain a posterior distribution.
Statistical model Created 2026-09-24 Updated 2026-11-05
A statistical model is a specified family of probability distributions for the observations. A statistical parameter often indexes it as , where is the parameter space. The choice states which patterns of variation are considered possible and supplies a basis for statistical inference; it does not guarantee that any member exactly describes the data-generating mechanism.
For example, suppose binary observations are independent and identically distributed with a common unknown success probability . The Bernoulli distribution model has and likelihood function
With and seven successes, this becomes and is maximized at . The binary observation type alone does not supply independence or a common : those are additional model assumptions. A regression model can instead let the response law depend on observed covariates, while a mixed model represents a specified grouping structure through random effects.
Sufficient statistic for a symmetric uniform sample Created 2026-10-06 Updated 2026-11-05
For independent samples uniform on with , the absolute maximum is a minimal sufficient statistic, since the likelihood is . The pair of sample minimum and maximum is a sufficient statistic but is not minimal: simultaneous sign reversal preserves and generally changes that pair.
An unbiased likelihood estimator has expectation equal to the likelihood function at each parameter. A pseudo-marginal algorithm requires this estimator to be nonnegative. Importance sampling provides for independent random variables , provided the proposal covers the target support. Relative variance affects the mixing time even though unbiasedness guarantees the correct marginal target.
Uniform-alternative binomial Bayes factor Created 2026-10-14 Updated 2026-10-25
For a binomial distribution count, compare a point null with the uniform alternative prior distribution on . The alternative Bayesian model evidence is
so the exact Bayes factor favouring the point null is . With standardized deviation , the local normal distribution approximation is . At fixed the frequentist p-value remains approximately constant, while the factor grows like . This demonstrates the prior-width mechanism behind the Jeffreys-Lindley paradox for a Gaussian point null in a discrete sampling model.
For a separate original example, change the point null to and take to have a uniform prior , fixed before the ten Bernoulli trials. Here is drawn once under and shared by all ten trials: they are independent conditional on . For one particular ordered sequence with eight heads and two tails, the Euler beta function gives
Thus the Bayes factor for the uniform alternative against the point null is
The binomial distribution count formulation gives and . Their ratio is again ; these count masses must not be combined with the ordered-sequence masses above. The general formula at the start of this article puts the point null in the numerator, so it equals here.
In contrast, maximum likelihood estimation gives : differentiating the log-likelihood yields , and the endpoints have zero likelihood. Its peak-to-null likelihood ratio is
This peak is chosen after seeing the data, so it is not the Bayes factor of the prespecified . The Bayesian model evidence for averages over its entire uniform prior; it is smaller than the peak. It is also smaller than . Against the same null, a point alternative fixed at would give a Bayes factor about , greater than . An integrated alternative need not outperform a point alternative.
Figure 1.
A uniform-prior average of the likelihood differs from its maximum
. Original calculation for a stipulated ordered sequence with eight heads and two tails. Left: the likelihood function and its uniform prior average , with the point null at and peak at . Right: the average, fixed value, and peak divided by the same null value . Only the first two comparisons use prespecified alternatives; the peak uses a maximum-likelihood estimate. These are calculated model quantities, not measured frequencies. Details.
For a general proper beta distribution prior distribution with , the same ordered-sequence marginal likelihood is . This follows by substituting the normalized prior density into the integral, and shows exactly where the parameter prior distribution changes the Bayes factor. Holding the parameter prior distributions fixed while changing model prior odds leaves that Bayes factor unchanged.
Uniform-Pareto model evidence Created 2026-10-07 Updated 2026-11-05
Let independent observations have uniform distributions on and let have a Pareto distribution with shape parameter and lower bound . For positive observations put . Integrating the likelihood function against the prior density gives the Bayesian model evidence
For independent groups with shared fixed hyperparameters, the Bayesian model evidences multiply. With groups and , the exact hyperparameter log-likelihood function is
The term may be omitted when optimizing with fixed. It must remain when varies: each then changes whenever exceeds that group's sample maximum. Calling this term a constant in a joint fit would change the maximum-likelihood estimator.
Unknown entry age in a Weibull survival segment Created 2026-10-06 Updated 2026-11-05
Conditional survival for a segment of length entered at age is under inverse-scale parameter . It depends on the entry age unless . Unknown ages require a specified entry-age model or recovered history; merely adding equally unspecified segments does not supply the missing likelihood information.
Weibull distribution Created 2026-09-24 Updated 2026-11-05
The two-parameter Weibull distribution is a positive lifetime model with shape and scale . Its Probability density function and Survival function are
Its Cumulative distribution function is . The Hazard function is : it decreases for , is constant for , and increases for . This permits decreasing or increasing failure rates within one family.[ref]
The mean is and the variance is . The substitution reduces the power-moment integrals to gamma integrals. At the law is exponential with rate ; at its survival is .
Inverse transform sampling is explicit: for uniform on , take . Conditional survival after an already attained age is , which depends on unless . This is why an Unknown entry age in a Weibull survival segment affects the likelihood. The scale parameter here is ; inverse-scale parameterizations instead use its reciprocal.
Zeros trick 2026-10-06
A zeros trick implements a positive likelihood function by adding an observed zero with Poisson distribution mean . If this mean is nonnegative throughout the parameter support, its likelihood is and yields the intended Bayesian posterior. The constant must be independent of the parameter; arbitrary clipping changes the target.