Parametric and Non-Parametric Statistics
Parametric and non-parametric statistics are methods used in the study of populations. Many empirical statistical researches have been carried out to determine whether parametric or non-parametric statistics should or should not be used in trials. Some books have condemned the use of parametric methods such as the t-tests, as some writers say that observations must be acquired from populations that are normally distributed.
But it is important to understand that, if any recommendation to the statistical methods should be based on the two types of error: Type I (inappropriate rejection of the null hypothesis), and Type II (inappropriate failure to reject the null hypothesis). However, the relative advantages of the two methods of statistical methods have been focused on the Type II error. This is because some studies that were performed showed that the t-tests carried out, rejected the null hypothesis most of the time.
A study of the Mann-Whitney and ANCOVA methods was carried out to state the comparison between the two methods. Vickers (2005) found out that ANCOVA (parametric method) out-powered Mann-Whitney (non-parametric method) in most circumstances. This study concluded that “There is no simple and obvious manner in which non-parametric methods becomes superior once the distribution of data shifts away from normal” (Vickers, 2005).
Vickers (2005) continues to say that when there is normality in the data, the parametric method is more efficient. According to Vickers (2005), there is no relationship that exists between the relative power and sample size. Data used in the parametric statistics is always interval or ration while the data to be used in the non-parametric statistics is ordinal or nominal.
It is therefore very clear that for parametric statistics, the observation must be independent whereas in non-parametric statistics, the observations are independent. Therefore, non-parametric are said to be less powerful since they utilize less information for the calculation process.
Reference Vickers, J. A. (2005). Parametric versus non-parametric statistics in the analysis of randomized trials with non-normally distributed data. Retrieved July 14, 2010 from: http://www. biomedcentral. com/1471-2288/5/35Sample Essay of Eduzaurus.com