A family of methods which makes no assumptions about the population distribution. Non-parametric methods most commonly work by ignoring the actual values, and, instead, analyzing only their ranks. This approach ensures that the test is not affected much by outliers, and does not assume any particular distribution. The clear advantage of non-parametric tests is that they do not require the assumption of sampling from a Gaussian population. When the assumption of Gaussian distribution does not hold, non-parametric tests have more power than parametric tests to detect differences.
Published in Chapter:
Non-Parametric Statistical Analysis of Rare Events in Healthcare: Case of Histological Outcome of Kidney Transplantation
Soheila Nasiri (University of Ottawa, Canada) and Bijan Raahemi (Unviersity of Ottawa, Canada)
Copyright: © 2017
|Pages: 15
DOI: 10.4018/978-1-5225-2515-8.ch009
Abstract
The assumption of Gaussian distribution of population does not always hold strongly in health studies. The sample size may not be large enough due to the limited nature of observations such as biopsies taken during kidney transplantation, the distribution of sample may not be Gaussian, or the observation may not even be possible for the far ends of a Gaussian distribution. In such cases, an alternative approach, called nonparametric tests can be applied. In this study, a non-parametric single center retrospective analysis of adult kidney transplant is performed to compare histological outcomes among three different groups of deceased kidney donors, based on the biopsies taken before and after kidney transplant at months 1, 3, and 12. A total of 107 transplants were observed in this study with 310 surveillance biopsy taken then classified based on the Banff 97 adequacy assessment. It is concluded that the recipient's internal condition after kidney transplant is as important as the donor's risk factors.