A thesis on ranking and selection in the method of paired comparison

Scariah, K S (1983) A thesis on ranking and selection in the method of paired comparison. PhD thesis, Indian Institute of Technology, New Delhi.

[img]
Preview
PDF
35Mb

Abstract

Experimental situations where the subjective judgements or appraisals of the individuals , lead to qualitative, comparative responses or situations where quantifications through measure ments are difficult or illusary, the method of paired comparison has a significant role to play. Interest in paired comparison has been generated through the paired nature of competitions of many kinds, through the experimental simplicit:y of sensory comparisons of items observed in pairs, and through combinatorial properties associated with the construction of experimental designs or tournaments. The method of paired comparisons has had a long and honourable history in psycological experiments. During the past three decades this method has attracted the attention of people from a wide spectrum of interest. In general, the method would be applicable whenever the objects can be presented to judges to assess their non-measurable quality in pairs either simulteniously or in succession. The basic goal in the method of paired comparison is to discriminate between t> 2 treatments, on the basis of preference data obtained by presenting the treatments in pairs in a given order to a set of judges according to a specified paired comparison design. In such experimental situations the method has led to a surprising amount of model building to provide simple stochastic representation. The main statistical problem with these models have generally been that of estimation of the true worths of the treatments, testing of goodness of fit of the model, testing the hypothesis of equality of treatments etc based on the score vector obtained by using a suitable economical paired comparison design. Another type of problem that is of interest is to select the best treatment with at least a specified high probability, using a definite selection rule. Selection rules are defined in terms of the score vector that is obtained by using the given design. A related problem i s that o f choosing a subset of given treatments which includes the best treatment with at least a specified high probability. These problems are respectively special cases of the two basic formulations referred to as Indifference zone formulation end Subset s e l ection formulation in the area of Ranking and Selection procedures for statistical populations.

Item Type:Thesis (PhD)
Supervisor(s):Sadasivan, G and Handa, B R
Uncontrolled Keywords:ranking and selection; method of paired comparison
Subjects:Theses
Divisions:CMFRI-Cochin > Marine Capture > Fishery Resource Assessment
ID Code:7082
Deposited By:Geetha P Mrs
Deposited On:27 Jan 2011 16:36
Last Modified:21 Jan 2014 12:32

Repository Staff Only: item control page