Suggested applying technique ( VAS ) for an Ordinal Category Data that could be transformed to quantitative measurement in Linear Discriminant analysis

	Suggested applying technique ( VAS ) for an Ordinal Category Data that could be transformed to quantitative measurement in Linear Discriminant analysis
AL-TAQANI
Article 1, Volume 22, Issue 5, 2009, Pages 0-0
Abstract
The research aimed to building a linear Discriminant function to classified the new individual through using two techniques , namely ( Original Data ) which depends on the whole data in the estimating the model function and by ( Cross – Validation ) which avoidance the observation that should be under identification in the building process and these was achieved by applying two methods : The First : Known by Conventional method , which might be confront the cases of included one ( or more ) categories variables (Ordinal Scale) measurement in the discriminate variables set that should be followed the multivariate normal distribution , and that surely causes increasing error of the classification in the Discriminant function . The Second : By estimation of the category response for these variables through using the suggested method known Visual Analogue Scales - (VAS) , which depends on the transformation the categories data that could be transformed to the interval scale . The superior conclusions that were obtained is the increasing of the coincidence ratio of classification of the new observation when included in the creating model and it was clearly decreased in the avoided state through applying the conventional technique ( before applying the transformation process – VAS ) and to perfect coincidence state for the classify the new observation when that observation included in the creating model after applying the transformation process to the category data that could be transformed and that was achieved by (VAS) . In addition to that relativity decreasing in the coincidence state when the avoided the new observation in the creating model process which indicate a highly importance for the transformation process through decreasing the classification error according to presented the multivariate normal distribution assumption and it's the most important assumptions that should be available in applying the linear Discriminant function .

Statistics Article View: 142 PDF Download: 23