nextuppreviouscontents
Next:Derivatives of system functionsUp:Estimating information and correlation Previous:Appendix F. Existence conditions

Appendix G. Derivatives of overlap convolutions: Poles

This appendix discusses derivatives with respect to tex2html_wrap_inline14461 of expressions such as tex2html_wrap_inline17773, where tex2html_wrap_inline17775, and tex2html_wrap_inline14461 may be any number within the constraints of existence. We consider the various cases that arise when some combination of poles occurs and demonstrate the various simplified expressions for the derivatives in these cases.

When tex2html_wrap_inline14461 is not an integer, there are no poles in either tex2html_wrap_inline17781 or tex2html_wrap_inline17783 and the usual derivative expressions hold. When tex2html_wrap_inline14461 is an integer and tex2html_wrap_inline17787, the usual expressions also hold since tex2html_wrap_inline17789 and therefore tex2html_wrap_inline17791. The case where the usual expressions hold will be denoted Case 0.

When tex2html_wrap_inline14461 is an integer and tex2html_wrap_inline16977, there are two cases are of interest. The first, Case 1, occurs when tex2html_wrap_inline16977 and tex2html_wrap_inline17803, so that there are poles in the denominator of only. The second, Case 2, occurs when tex2html_wrap_inline16977 and tex2html_wrap_inline17809, so that there are poles in both the numerator and the denominator.

In order to find expressions for the derivatives in cases Case 1 and Case 2 we use the following three facts. 1) The only singularities of the gamma function tex2html_wrap_inline17815 are simple poles at tex2html_wrap_inline17817 with residues tex2html_wrap_inline17819 respectively. 2) tex2html_wrap_inline17821 whenever the expressions exist. (The identity tex2html_wrap_inline17823 may be used in deriving this.) 3) tex2html_wrap_inline17825 is the representation (away from the poles in the gamma functions) of an analytic function (note that tex2html_wrap_inline17789 is still assumed). Using these facts, the expressions for Case 1 and Case 2 are found by substituting tex2html_wrap_inline17833 for tex2html_wrap_inline14461 (now restricted by the conditions of Case 1 and Case 2 to be a nonnegative integer) in the corresponding case Case 0 expressions and taking the limit tex2html_wrap_inline17549.

Case 0tex2html_wrap_inline14461 non-integer, or tex2html_wrap_inline14461 an integer with tex2html_wrap_inline17787 and tex2html_wrap_inline17803. There are no poles in the numerator or denominator of tex2html_wrap_inline17825.

The first derivative is given by
eqnarray9238

The rth derivative may be found by iteration, using equation 9.141 and the recursion relation
eqnarray9243

The second derivative is given by
eqnarray9245

Case 1tex2html_wrap_inline14461 an integer, tex2html_wrap_inline16977 and tex2html_wrap_inline17803. The denominator contains a pole.

The zeroth derivative is 0.

Taking the appropriate limit in equation 9.141 gives us the first derivative
eqnarray9253

Taking the appropriate limit in equation 9.143 yields the second derivative
eqnarray9255

Case 2tex2html_wrap_inline14461 an integer, tex2html_wrap_inline16977 and tex2html_wrap_inline17809. Both the numerator and denominator contain poles.

The zeroth derivative is simply
equation9257

Taking the appropriate limit in equation 9.141 gives the first derivative
equation9259

Taking the appropriate limit in equation 9.143 yields the second derivative
equation9264


nextuppreviouscontents
Next:Derivatives of system functionsUp:Estimating information and correlation Previous:Appendix F. Existence conditions
David Wolf

Tue Mar 25 08:11:49 CST 1997