01.05.2017 Views

563489578934

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

B–9 Multivariate Statistics 713<br />

x 1<br />

y 1<br />

x<br />

x 2<br />

x 3<br />

•••<br />

y = h(x)<br />

Transfer characteristic<br />

(no memory)<br />

y 2<br />

y 3<br />

•••<br />

y<br />

x N<br />

y N<br />

Figure B–13<br />

Multivariate functional transformation of random variables.<br />

THEOREM. Let y = h(x) denote the transfer characteristic of a device (no memory)<br />

that has N inputs, denoted by x = (x 1 , x 2 ,..., x N ); N outputs, denoted by y = (y 1 , y 2 ,..., y N );<br />

and y i = h i (x). That is,<br />

y 1 = h 1 (x 1 , x 2 , Á , x N )<br />

y 2 = h 2 (x 1 , x 2 , Á , x N )<br />

o<br />

y N = h N (x 1 , x 2 , Á , x N )<br />

(B–98)<br />

Furthermore, let x i , i = 1, 2,..., M denote the real roots (vectors) of the equation y = h(x). The<br />

PDF of the output is then<br />

f y<br />

(y) = a<br />

M<br />

i=1<br />

f x (x)<br />

|J(y>x)|<br />

`<br />

x = x i = h i -1 (y)<br />

(B–99)<br />

where |·| denotes the absolute value operation and J(y/x) is the Jacobian of the coordinate<br />

transformation to y from x. The Jacobian is defined as<br />

0h 1 (x)<br />

0x 1<br />

0h 2 (x)<br />

J¢ y x ≤ = Det H 0x 1<br />

o<br />

0h N (x)<br />

0x 1<br />

0h 1 (x)<br />

0x 2 p<br />

0h 1 (x)<br />

0x N<br />

0h 2 (x) 0h 2 (x)<br />

0x 2 p 0x N X<br />

o<br />

o<br />

0h N (x)<br />

p 0h N(x)<br />

0x 2 0x N<br />

(B–100)<br />

where Det[·] denotes the determinant of the matrix [·].<br />

A proof of this theorem will not be given, but it should be clear that it is a generalization<br />

of the theorem for the one-dimensional case that was studied in Sec. B–8. The coordinate transformation<br />

relates differentials in one coordinate system to those in another [Thomas, 1969]:<br />

dy 1 dy 2<br />

Á dy N = J¢ y x ≤ dx 1 dx 2<br />

Á dx N<br />

(B–101)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!