nextuppreviouscontents
Next:Estimators for the first Up:Estimating functions of probability Previous:Form of the integrals

Integrating

The results of this section are presented as a series of theorems leading quickly to the estimators of the moments of the entropy, presented in the next section. Note that the conditions of existence of the results are given in full generality in the complex domain, although the actual event counts are non-negative integers.

Define the Laplace convolution operator tex2html_wrap_inline16329 by its operation on two functions
equation5440

. If tex2html_wrap_inline16331 then
equation5445

Proof: Define tex2html_wrap_inline16341. Define tex2html_wrap_inline16343. Define tex2html_wrap_inline16345. The integral in the statement of the theorem, after substituting for tex2html_wrap_inline16347 and tex2html_wrap_inline16349, becomes
eqnarray5455
Integrating over tex2html_wrap_inline16351 the integral takes the form
eqnarray5457
The inner integral may be written as a convolution, and noting that tex2html_wrap_inline16353 must be positive we find
eqnarray5459
Clearly, this can be extended by induction and the associativity of the convolution operator to give the desired result. QED.

For the following, the Laplace transform operator L and its inverse are discussed in appendix 9.6.3.

. (Laplace convolution) If tex2html_wrap_inline16357 exists for tex2html_wrap_inline13113, then
equation5461

Proof: The proof is lengthy, but may be found in [36]. The result is mentioned in [50, 54].

Define the Gamma function tex2html_wrap_inline16361tex2html_wrap_inline16363 by
equation5465

. If tex2html_wrap_inline16365tex2html_wrap_inline13113, then
equation5467

Proof: By theorem 1
equation5472
By theorem 2
equation5475
Note that
equation5479
Substitute from equation 9.27 in the right side of equation 9.26. Finally use equation 9.27 again to take the inverse. Setting tex2html_wrap_inline16373 finishes the proof. QED.

. If tex2html_wrap_inline16375tex2html_wrap_inline13113, then
equation5481

Proof: Note that tex2html_wrap_inline16383. The interchange of derivative and integral is given in appendix 9.7.

In using theorem 4 we will take the derivatives in the expression on the right side of equation 9.28 where N appears in tex2html_wrap_inline16387 (see equation 9.24); thus note that since tex2html_wrap_inline16389, we have tex2html_wrap_inline16391.

Define the polygamma function
equation5492
(see [3] for properties of the polygamma function), and define
equation5494
Introduce
equation5496

. For tex2html_wrap_inline16375tex2html_wrap_inline13113,
eqnarray5503

Proof: By theorems 3 and 4
eqnarray5508
Taking the derivative and substituting for the tex2html_wrap_inline16403 function yields the result. QED.

. For tex2html_wrap_inline16375tex2html_wrap_inline13113,

(6.1)
eqnarray5514

(6.2)
eqnarray5527

Proof: Apply theorems 3 and 4 in a manner similar to the proof of theorem 5. Here two derivatives are needed.


nextuppreviouscontents
Next:Estimators for the first Up:Estimating functions of probability Previous:Form of the integrals
David Wolf

Tue Mar 25 08:11:49 CST 1997