Record Details

A STUDY OF TWEEDIE FAMILY OF DISTRIBUTIONS FOR RAINFALL MODELLING

KrishiKosh

View Archive Info
 
 
Field Value
 
Title A STUDY OF TWEEDIE FAMILY OF DISTRIBUTIONS FOR RAINFALL MODELLING
M.Sc.
 
Creator LALL, SHWETANK
 
Contributor Ghosh, Himadri
 
Subject marketing, precipitation, sampling, byproducts, paper, biological phenomena, vegetables, forecasting, self help, wells
 
Description The present study deals with the investigation of behaviour of monthly rainfall data.
There are many exact zeroes in the rainfall recording of a month over a period of
years. Probability distributions currently in use are not able to combine the discrete
aspect of getting an exact zero reading of rainfall in a month and the aspect of amount
of rainfall in the same month. Compound Poisson Tweedie distribution, however
provides a better way to combine both the aspects in one probability distribution. In
this work the compound Poisson Tweedie distribution has been used to fit the
monthly rainfall data by computing the estimates of its parameters. For the purpose of
estimation of parameters of compound Poisson Tweedie distribution a new algorithm
has been developed and used in this work. Another approach to model monthly
rainfall is modified gamma distribution with singularity at zero. This is a three
parameter mixture distribution which gives a probability mass to the occurrence of
zero in a month and defines a conditional density for the amount of rainfall. The
parameters are estimated by developing an algorithm. For testing the goodness of fit
of the distributions Kolmogrov-Smirnov test is used. Since the test requires the
complete continuity of the distributions in their domain, we computed the theoretical
distribution function both the distribution by using trapezoidal rule of numerical
integration. The test shows that both the distributions are good for fitting the monthly
rainfall data. In this investigation bootstrap technique has been employed to know the
distribution of Kolmogrov-Smirnov test statistic. The bootstrap technique provides the
relative goodness of fit of the proposed distribution. Thus it can be concluded that the
Tweedie family of distributions are good for fitting rainfall months with few or many
zero readings
 
Date 2016-03-10T17:02:23Z
2016-03-10T17:02:23Z
2013
 
Type Thesis
 
Identifier http://krishikosh.egranth.ac.in/handle/1/65054
 
Language en_US
 
Format application/pdf
 
Publisher IARI, IASRI, NEW DELHI