On a common positive support, a factorization $f={\varphi }_0(x){\varphi }_C(x;\psi ){\varphi }_S(x;\lambda )$ implies ${\partial }_{\psi }{\partial }_{\lambda }\log f=0$. Thus any nonzero mixed derivative prevents this exact separation. For $n$ independent gamma distributed observations with shape $a$ and rate $b$, this mixed derivative is $n/b$, so exact separation of shape and rate into such factors is impossible.

The Heyting implication defines a nucleus on a frame whose sublocale is the open determined by $a$. It is the least nucleus on a frame sending $a$ to $1$.

This nucleus on a frame determines the closed sublocale complementary to the open determined by $a$. It is the least nucleus on a frame taking $0$ to $a$. In the frame of nuclei on a frame, $c(a)$ and the open nucleus $o(a)$ are complements.

For a frame homomorphism $h:B\to A$, nuclei are equivalent when their restrictions to $h(B)$ agree. The least representative of the class of $j$ isThe least representatives form a frame generated by $c(A)$ and $o(h(B))$. To see minimality, each summand is below $j$ and its value at $hb$ is $j(hb)$; any equivalent nucleus therefore bounds every summand. This construction gives the pushout of $B\to N(B)$ along $h$.

A translation-invariant connected generating functional with a constant external source is extensive in spacetime volume ${\mathcal{V}}_d$. Its total logarithm is dimensionless, while its density has mass dimension $d$. Consequently its dimensionless representation at fixed physical volume must retain ${\mu }^d{\mathcal{V}}_d$. For a free massive scalar with source coupling ${\mu }^{-ϵ/2}h\varphi$, completing the square gives $F={\mathcal{V}}_d{\mu }^{-ϵ}{h}^2/(2m^2)$. This already disproves a literal rewriting of the full uniform-source logarithm using only $\lambda ,m^2/{\mu }^2,h/{\mu }^3$ at fixed volume. In infinite volume the total functional is generally infinite; use the density, a source of compact support, or an explicit volume regulator. After writing the density as ${\mu }^{d}f$, its renormalization group equation has the engineering term $+df$. Ultraviolet subtraction alone does not establish infrared finiteness for a massless constant source.

Because $f(z)/z$ is holomorphic and nowhere zero on the unit disc, its square root with value one at zero exists. Defining $h(z)=z\sqrt{f(z^2)/z^2}$ gives an odd normalized univalent function. If $h(z)=h(w)$, squaring gives $z^2=w^2$; the possibility $z=-w$ and oddness force both to be zero. Applying the area theorem for univalent functions to $1/h(1/\zeta )$ yields $|a_2|\le 2$.

The odd square-root transform of a normalized univalent function has reciprocal exterior expansion $1/h(1/\zeta )=\zeta -(a_2/2){\zeta }^{-1}+\cdots$. Its first negative Laurent coefficient has modulus at most one by the area theorem for univalent functions, proving the bound. If $w$ is omitted by $f$, the normalized map $wf/(w-f)$ has second coefficient $a_2+1/w$. Applying the bound twice gives $|1/w|\le 4$, the Koebe quarter theorem.

A finite group is p-nilpotent when it has a normal p-complement. This terminology is distinct from having a normal Sylow p-subgroup. A finite group is nilpotent exactly when it is p-nilpotent for every prime dividing its order.

An antinode is a position of maximum oscillation amplitude in a standing wave. For $u_n=A\sin (n\pi x/L)\cos ({\omega }_{n}t+\phi )$, the antinodes occur atwhere the amplitude is $|A|$. They sit halfway between consecutive nodes of a standing wave. For the first normal mode, the sole antinode is $x=L/2$. It reaches zero displacement at a quarter oscillation period while moving fastest, which illustrates why an antinode is defined by oscillation amplitude rather than instantaneous height.

A node of an ideal standing wave is a position whose oscillation amplitude is zero: its displacement stays zero for all time. For a nontrivial normal mode $u(x,t)=\varphi (x)q(t)$, this means $\varphi (x)=0$. A momentary zero caused by $q(t)=0$ does not make every position a node. In more dimensions the zeros of a mode shape can form curves or surfaces.

For the $n$th normal mode of a uniform fixed-end wave equation on a string, the nodes of a standing wave areThere are $n+1$ nodes of a standing wave including the two endpoints, and $n-1$ interior ones. For $n=3$ and $L=1\mathrm{m}$, they lie at $0$, $1/3$, $2/3$, and $1\mathrm{m}$. Each remains stationary through every snapshot in the normal mode diagram; an antinode instead marks a maximum oscillation amplitude. ^[ref]

For the undamped first normal mode of a fixed string, start at maximum displacement with $u(x,t)=A\sin (\pi x/L)\cos ({\omega }_{1}t)$. In the small-slope wave equation on a string, its kinetic energy and excess potential energy areThe potential energy comes from the extra length under tension: $\sqrt{1+u_x^2}-1\simeq u_x^2/2$. It depends on slope, so a node of a standing wave need not have zero local potential energy density. These are physical energies, with the mass and tension factors restored to the normalized wave energy. ^[ref]

Integrating ${\sin }^2(\pi x/L)$ and ${\cos }^2(\pi x/L)$ over the string gives $L/2$ in both cases. Using $\mu {\omega }_1^2=\tau {\pi }^2/L^2$ yieldsFor the model $\mu =1\mathrm{kg}{\mathrm{m}}^{-1}$, $\tau =1\mathrm{N}$, $L=1\mathrm{m}$, and $A=0.010\mathrm{m}$, the angular frequency is ${\omega }_1=\pi \mathrm{rad}{\mathrm{s}}^{-1}$, the frequency is $0.5\mathrm{Hz}$, and $E\simeq 2.4674\times 10^{-4}\mathrm{J}$. The maximum slope $A\pi /L\simeq 0.0314$ is small. At $t=0$ all the energy is potential energy; at $t=T_1/4=0.5\mathrm{s}$ the string is straight and all the energy is kinetic energy. At $t=T_1/2=1\mathrm{s}$ the displacement is reversed and all the energy is again potential energy. The energy curves repeat twice during one displacement cycle, while conservation of energy keeps their sum constant.

Distinct squared traction-free eigenfrequencies in a bounded lossless elastic body have eigenfunctions orthogonal in $L^2(\mathcal{D},\rho dV{)}^3$, by the Betti identity for elastodynamic fields. Normalize one field from each eigenspace. This is an orthonormal set in a separable Hilbert space, so it is countable: disjoint balls of radius less than $1/\sqrt{2}$ around its unit vectors must contain distinct points of a countable dense set. Positive and negative frequency partners add at most a factor of two. This argument proves countability without requiring a separate compact-resolvent proof.

The mode shape is the fixed spatial eigenfunction multiplying the temporal factor of a normal mode. It is determined by the governing differential equation and boundary conditions, and may be normalized by a chosen amplitude or energy convention.

For the same normal distribution model, a two-sided test compares the null hypothesis $H_0:\mu ={\mu }_0$ with the alternative hypothesis $H_1:\mu \ne {\mu }_0$. Under the null hypothesis, the standardized test statistic has standard normal distribution. Write $\Phi$ for the standard normal distribution function. With $\bar{x}=1.2$, $\sigma =2$, $n=16$ and ${\mu }_0=0$, the observed test statistic is $z=1.2/0.5=2.4$. Its p-value isThus the p-value is a tail probability computed under the null hypothesis, including departures in both directions. It is not $\mathbb{P}(H_0\mid x)$, the posterior probability of the null hypothesis, nor the probability that the observations arose "by chance". The normal distribution test statistic and its calibration appear in NIST's teaching reference for testing a process mean; the tail interpretation is stated in Penn State's hypothesis-testing notes.

Choose the significance level $\alpha =0.05$ before observing the sample. The rejection region is $|Z|>z_{0.975}$, whose probability under $H_0$ is exactly 0.05. Here $2.4>1.9599639845\dots$, so reject $H_0$. This agrees with zero lying outside $[0.22,2.18]$. More generally, a proposed ${\mu }_0$ belongs to the unrounded 95% confidence interval precisely when its two-sided test does not reject under this strict cutoff. The equality boundary has probability zero in this continuous statistical model.

The significance level and statistical power describe different repeated-experiment probabilities. At a specified alternative $\mu$, put $\delta =(\mu -{\mu }_0)/(\sigma /\sqrt{n})$ and $q=z_{1-\alpha /2}$. ThenThe first is the Type I error probability; the second is the Type II error probability at that alternative. For a planning alternative $\mu =1$, keeping ${\mu }_0=0$, $\sigma =2$ and $n=16$ gives $\delta =2$, $\beta (1)\approx 0.483995$ and statistical power $0.516005$. At $n=64$, $\delta =4$ and statistical power rises to $0.979327$. These are model calculations for specified alternatives, not posterior probabilities inferred from a nonsignificant result. They explain why failure to reject cannot establish equality and why choosing sample size before an experiment is useful.

Historically, Ronald Fisher's 1925 book discusses the normal cutoff $1.96$ and tail probability 0.05 as a convenient convention on page 47, rather than a mathematically compulsory threshold: ^[ref].

For a normal matrix, $N^{†}N=N{N}^{†}$. Consequently $\Vert Nx{\Vert }^2=x^{†}{N}^{†}Nx=x^{†}N{N}^{†}x=\Vert N^{†}x{\Vert }^2$. In particular, $N$ and its Hermitian conjugate have the same kernel of a linear map.

With repeated responses at an identical predictor configuration, pure error is their within-configuration variability. Its residual sum of squares can estimate measurement variance independently of a candidate regression mean function. Without replication, a fitted residual can combine pure error and lack of fit.

Lack of fit is a mismatch between a regression model's mean function and the true mean relationship. Replicated predictor configurations can separate its sum of squares from pure error. A residual-versus-fitted plot can suggest curvature or changing variance even without a formal replication-based test.

For a full-column-rank design $X$ and $Y=X\beta +\epsilon$ with $\epsilon \sim N_n(0,{\sigma }^{2}I)$, the maximum-likelihood estimators are $\hat{\beta }=(X^{T}X{)}^{-1}{X}^{T}Y$ and ${\hat{\sigma }}^2=\Vert Y-X\hat{\beta }{\Vert }^2/n$. Orthogonal Gaussian projections give their displayed distributions and independence. The variance maximum-likelihood estimator has expectation $(n-p){\sigma }^2/n$; the unbiased residual variance divides by $n-p$ instead.

Analysis of covariance combines a categorical treatment and a quantitative covariate in a normal linear model. A common-slope model is $Y_i=\alpha +\beta x_i+{\sum }_{g=2}^G{\tau }_g{1}_{\{G_i=g\}}+{\epsilon }_i$, with independent $N(0,{\sigma }^2)$ errors. Treatment differences compare fitted means at the same covariate value. Adding treatment-by-covariate interactions permits separate slopes, and a partial F-test compares this extension with the parallel-lines model. The treatment test after adding the covariate differs from an unadjusted comparison, because treatment groups can have different covariate distributions.

For a full-column-rank design matrix in a normal linear model, a fitted mean at covariate vector $v$ is $v^T\hat{\beta }$. The covariance matrix of the least-squares estimator immediately gives the displayed variance. Replacing ${\sigma }^2$ by the independent unbiased residual variance gives its squared standard error and a Student t confidence interval on $n-p$ statistical degrees of freedom. A future response instead contributes an additional error variance ${\sigma }^2$ when its error is independent of the fitted data.

// Target: probability-and-statistics.bigb