Discordant observations in flood frequency analysis

Abstract

The discordant observation is an observation that is far away from the rest of the data set and newlinein the case of flood frequency analysis; it defined as the observation which is far away from newlinethe rest of the flood events. As it had observed that the presence of discordant could mislead newlinethe analysis, no issue whatever is the study. But dealing with the data set/ experiment which is newlinedirectly associated with the commons, makes it a more serious matter. Thus, the detection of newlinediscordant observation in the flood frequency analysis becomes more essential as the field is newlinedirectly associated with the commons. Thus, in this thesis, our main focus is first to suggest a newlinetesting technique to detect the discordant i.e. to define a technique that will help identify the newlinediscordant. In this process, its effect has been observed in the parameter estimates obtained by newlinedifferent estimation techniques namely the least square method, maximum likelihood method, newlineand method of moments. These parameter estimates were obtained for different distributions newlinelike Normal distribution, Pearson type III distribution, Log normal distribution, Exponential newlinedistribution, Gumbel distribution, and Weibull distribution, and then after three cases were newlineconsidered (and#119894;) without removing discordant, (and#119894;and#119894;) with removing single discordant and (and#119894;and#119894;and#119894;) newlineremoving multiple discordant observations and also the percentage changes observed between newlinethese cases. newlineThrough this study, it had found that the change in the shape parameter reflects the most. newlineThis result becomes the motivation for the introduction of a new testing procedure, which newlineinvolves the third parameter. In the next chapter, a new test statistic, which involves the third newlineparameter of the Weibull distribution, was proposed. For this study, Weibull samples up to size newline50 were generated by fixing the values of the location parameter and scale parameter and newlinechanging the values at 1, 2, and 3 for the third parameter and thus, obtained critical value and#945; at newline1%, 5 % and 10 % level of significa