Haro Stettner

Pelvic lymphadenectomy in operative treatment of ovarian cancer

From the end of 1979 to September, 1985, radical pelvic lymphadenectomy was performed at the Graz Clinic in 123 cases of Stages IA to IV ovarian cancer following maximum debulking procedure. In 97 patients lymphadenectomy was done primarily. In 26 it was performed during a follow-up operation to chemotherapy. The frequency of pelvic node involvement was 61.8% in the total material and 78.0% in 82 cases of Stage III disease only; 75.0% positive nodes were found in Stage III after chemotherapy. Aortic nodes were positive in 41.4%, but only when pelvic nodes were also positive. The 5-year actuarial survival rate for Stage III disease was 53.0% after pelvic lymphadenectomy compared with 13.0% without. In cases with negative nodes the survival rate was 74.7%; with positive nodes the survival rate was 45.9%.

Backfitting and Related Procedures for Non-parametric Smoothing Regression: A Comparative View

In this paper we relate backfitting (FRIEDMAN and STUETZLE, 1981) to the standard iterative procedures, i.e. Jacobi and Gauss-Seidel, developed for solving linear equation systems with non-singular system matrices. Aspects of improving the performance of these methods through relaxation are also considered. When fitting additive regression models non-parametrically by linear scatterplot smoothers, we are confronted with a singular system matrix of a certain block structure. Backfitting commonly applied (BIJA, HASTIE and TIBSHIRANI, 1989), although not designed for this situation, is examined. Because of a lack of theory behind this application and unknown convergence behaviour a simulation experiment is carried out applying cubic smoothing splines in a simple additive model. The behaviour of all three iterative procedures is studied by comparing the obtained results with those from a non-iterative Tichonow method (SCHIMEK, STETTNER and HABERL, 1992). The Jacobi iteration turns out to be superior to the other procedures when speed and precision are considered at the same time. Relaxation is of minor importance in Jacobi iteration.