next up previous contents
Next: Appendix C. Transforms Up: Estimating information and correlation Previous: Appendix A. Hypergeometric functions

 

Appendix B. Hypergeometric function identities

. An alternate proof of theorem 12 leads to an identity. Where the inverse T transform is applied in the proof of theorem 12, instead find the convolution tex2html_wrap_inline17309 using theorem 10.2 and express it in terms of tex2html_wrap_inline16789. Now, do the inverse transform and equate this result to the result of theorem 12 to find Gauss's identity:
 equation8955
. An alternate proof of theorem 14a leads to another identity. Instead of applying theorem 11, apply theorem 10.1 to the first two terms of the pairwise overlap convolution tex2html_wrap_inline17315 and immediately take the inverse tex2html_wrap_inline17317 transform. Now, do the final convolution with tex2html_wrap_inline17319 and take the inverse tex2html_wrap_inline17321 transform. Note that there is only a single summation in the result, whereas in the result in theorem 14a there are two summations. On the other hand, the convolution tex2html_wrap_inline17323 can be taken first, followed by a convolution with tex2html_wrap_inline17325, effectively interchanging indices 1 and 2. Equating these two single-sum forms gives the identity
 eqnarray8969
while equating either of the single-sum results just described to the original result of theorem 14a yields Gauss's identity equation 9.120, above. (To make the symmetry of equation 9.121 obvious, try substituting tex2html_wrap_inline17331).

. Utilizing Gauss's identity (see equation 9.120 above and [3], equation 15.1.1) provides further simplification in theorem 15a for cases 15a.2 and 15a.4. These simplifications are due to simplifications appearing in tex2html_wrap_inline17055 and tex2html_wrap_inline17025 respectively. The choice of the form of the results presented was made considering the simplicity of the results and consistency between the results.



David Wolf
Tue Mar 25 08:11:49 CST 1997