Zonal polynomials on monomial symmetric functions

吴炳洋

已于 2022-02-20 11:03:44 修改

阅读量188

点赞数

文章标签：数字通信统计学矩阵

于 2022-02-19 21:21:01 首次发布

本文链接：https://blog.csdn.net/wubyatseu/article/details/123017977

版权

If $\kappa =(k_1,k_2,\cdots,k_m)$ , the monomial symmetric function of $y_1,y_2,\cdots, y_m$ corresponding to
$\kappa$ is defined as

$M_\kappa(Y)=\sum ... \sum y_{i1}^{k_1}y_{i2}^{k_2}\cdots y_{im}^{k_m} =y_1^{k_1}y_{2}^{k_2}\cdots y_m^{k_m}+\textrm{symmetric terms}$ (1)

Zonal polynomials can be expressed in terms of the monomial symmetric functions:

$C_\kappa(Y)=\sum _{\lambda \le \kappa} c_{\kappa, \lambda } M_\lambda (Y))$ (2)

where the $c_{\kappa, \lambda }$ are constants and the summation is over all partitions $\lambda$ of k with $\lambda \le \kappa$ (that is, $\lambda$ is below or equal to $\kappa$ in the lexicographical ordering).

Substituting this expression (2) in the partial differential equation

$\Delta _Y C_\kappa (Y) =\left [ \rho_\kappa +k(m-1) \right ]C_\kappa (Y)$ (3)

and equating coefficients of like monomial symmetric functions on both sides leads to a recurrence relation for the coefficients, namely,

$c_{\kappa, \lambda }=\sum _{\lambda <\mu \le \kappa}\frac{\left [ (l_i+t)-(l_j-t) \right ]}{\rho_\kappa -\rho_\lambda } c_{\kappa, \mu}$ (4)

where $\lambda =(l_1,\cdots l_i,\cdots,l_j,\cdots, l_m)$ and $\mu =(l_1,\cdots l_i+t,\cdots,l_j-t,\cdots, l_m)$ . Such that, when the parts of the partition $\mu$ are arranged in descending order, $\mu$ is above $\lambda$ and below or equal to $\kappa$ in the lexicographical ordering.
The summation in (4) is over all such $\mu$ , including possibly, nondescending ones, and any empty sum is taken to be zero.

The partial differential operation on $M_{\mu}(Y)$

$\Delta _Y M_\lambda (Y) =\left [ \rho_\lambda +k(m-1) \right ]M_\lambda (Y)+\textrm{lower weight terms}$ (5)

produces

(a) $\rho_\mu +k(m-1) \hspace{2em} M_\mu (Y)$ and

(b) $l_i-l_j+2t \hspace{4em} M_ \lambda(Y)$ ,

the $l_i-l_j+2t$ is $(k_i-k_j)$ in $\mu$ actually.

eg.

$\Delta _Y M_{(4)} (Y) =\left [ \rho_{(4)} +4(m-1) \right ]M_{(4) }(Y)+4M_{(3,1) }(Y)+4M_{(2,2) }(Y)$

From $c_{(k),(k)}=1$ , the other coefficients can be obtained according to the recurrence relation.

eg.

$\large \begin{align*} c_{(4),(4)}&=1\\ c_{(4),(4)}&\rightarrow c_{(4),(3,1)}\\ c_{(4),(4)} , c_{(4),(3,1)}&\rightarrow c_{(4),(2,2)}\\ c_{(4),(3,1)}, c_{(4),(2,2)} &\rightarrow c_{(4),(2,1,1)}\\ c_{(4),(2,1,1)} &\rightarrow c_{(4),(1,1,1,1)}\end{align*}$

To get the partion (2,1,1), the partion (3,1) include (3,1,0) and (3,0,1) ; To get the partion (1,1,1,1), the partion (2,1,1) include (2,1,1,0) ,(2,1,0,1),(2,0,1,1)(1,2,1,0)(1,2,0,1) and (1,1,2,0).

Corollarys:

1. Let the coefficients $\large c_{\kappa,\lambda }$ be given by (4) and suppose that $\large \kappa$ is a partition of k into p nonzero parts. If the partition $\large \lambda$ of k has less than p nonzero parts and $\large \lambda <\kappa$ then $\large \large c_{\kappa,\lambda }=0$ .

2. If the $\large m\times m$ symmetric matrix Y has rank r , so that $\large y_{r+1}=\cdots = y_{m} =0$ y, and if $\large \kappa$ is a partition of k into more than r parts, then $\large C_\kappa (Y)=0$ .

3. If Y is a positive definite matrix ( Y > 0) then $\large C_\kappa (Y)>0$

If the partition $\large \kappa$ of k has $l(\kappa )$ nonzero parts, the value of the zonal polynomial at $\large I$ , is given by (an alternative notations for the zonal polynomials):

$\large Z_{\kappa}\left(\mathbf{I}\right)=|\kappa|!\,2^{2|\kappa|}\,{\left[m/2\right]_ {\kappa}} \frac{\prod\limits_{1\leq j<l\leq\ell(\kappa)}(2k_{j}-2k_{l}-j+l)}{ \prod\limits_{j=1}^{\ell(\kappa)}(2k_{j}+\ell(\kappa)-j)!}$

where

$\large \left | \kappa \right |=k$

$\large {\left[a\right]_{\kappa}}=\frac{\Gamma_{m}\left(a+\kappa\right)}{\Gamma_{m} \left(a\right)}=\prod_{j=1}^{m}{\left(a-\tfrac{1}{2}(j-1)\right)_{k_{j}}}$

.(DLMF: 35.4 Partitions and Zonal Polynomials)

A. G. Constantine. Some Non-Central Distribution Problems in Multivariate Analysis

See Muirhead (1982, pp. 68–72) for the definition and properties of the Haar measure d⁢H. See Hua (1963, p. 30), Constantine (1963), James (1964), and Macdonald (1995, pp. 425–431) for further information.

Alternative notations for the zonal polynomials are Cκ⁡(T) (Muirhead (1982, pp. 227–239)), 𝒴κ⁡(T) (Takemura (1984, p. 22)), and Φκ⁡(T) (Faraut and Korányi (1994, pp. 228–236)).

Reference： Aspects of multivariate statistical theory

吴炳洋

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
Zonal polynomials on monomial symmetric functions

Zonal polynomials can be expressed in terms of the monomial symmetric functions, and the coefficients can be obtained according to a recurrence relation.
复制链接

扫一扫