Exact finite-sample geometry of the Jarque-Bera statistic: support, discriminants and elliptic integrals

Under Gaussian sampling, removing location and scale turns the sample into a direction that is uniformly distributed on a sphere of dimension $n-2$, and sample skewness and kurtosis become a cubic-quartic polynomial image of spherical measure. We use this geometry to obtain exact finite-sample results for the Jarque-Bera statistic $J{B}_n$. First, for every $n\ge 3$ the largest attainable value of $J{B}_n$ is $n\{(n-2{)}^4+4(n-1{)}^2\}/\{24(n-1{)}^2\}$, attained by a single-outlier sample, so the asymptotic $10$ and $5$ tests have size zero for $n\le 6$ and the $1$ test for $n\le 7$; near this maximum the density of $J{B}_n$ behaves like an explicit multiple of $(J_n^{+}-x{)}^{(n-4)/2}$. Second, passing from residual coordinates to power sums writes the joint density of skewness and kurtosis as an integral of the reciprocal square root of a polynomial discriminant; for every $n\ge 5$ the innermost integral runs over a single interval and is a Lauricella $F_D$ period. Third, this yields explicit laws: an arcsine law for $n=3$, an algebraic joint density for $n=4$, and a single complete elliptic integral, equivalently ${}_2{F}_1(\frac{1}{2},\frac{1}{2};1;\cdot )$, for $n=5$. Residual-coordinate collisions generate the discriminant singularities of the joint law, while stationary points of $J{B}_n$ on the residual sphere govern the singularities of its one-dimensional density.