Selection of Feature Variables in Discriminant Analysis

<p>The auditing profession is at a crossroad worldwide. It currently faces many challenges especially in Greece where auditor rotation has been instituted as mandatory by EU regulation and the auditing profession is going to be fully liberalized (no limits on audit fees) . Given the Greek environment, it is important to investigate how client companies select auditors. In this study we address three questions. First, can selection of auditors be forecasted? Second, which statistical technique better fits the data set? Third, are there differences in firms’ financial ratios as well as institutional factors that affect auditor choice? Clients’ selection of auditors is considered in a research context using discriminant analysis and logistic regression. The discriminating factors between the two groups of companies include some firm financial ratios and institutional factors: QATA(Quick Assets/Total Assets) when using one year data, and QATA(Quick Assets/Total Assets) and SHAREHOLD (level of shareholdings) when using two year data. Prediction accuracy is close to 60.0 percent using discriminant analysis and around 80.0 percent using logistic regression. The contribution of this study is that it discriminates between the two groups of companies (Big Four versus second-tier or local auditing firms) in an IFRS environment.</p>

Download Full-text

Immunological Diagnosis of Human Cystic Echinococcosis: Utility of Discriminant Analysis Applied to the Enzyme-Linked Immunoelectrotransfer Blot

Clinical and Diagnostic Laboratory Immunology ◽

10.1128/cdli.6.4.504-508.1999 ◽

1999 ◽

Vol 6 (4) ◽

pp. 504-508 ◽

Cited By ~ 8

Author(s):

I. Gadea ◽

G. Ayala ◽

M. T. Diago ◽

A. Cuñat ◽

J. García de Lomas

Keyword(s):

Discriminant Analysis ◽

Hydatid Cyst ◽

Hydatid Disease ◽

Multivariate Technique ◽

Immunological Diagnosis ◽

History Of ◽

Pulmonary Hydatid Cyst ◽

Human Cystic Echinococcosis ◽

Selection Of

ABSTRACT An enzyme-linked immunoelectrotransfer blot for the diagnosis of human hydatid disease was performed, and the different antibody responses were analyzed by a discriminant analysis. This multivariate technique gave us, first, a selection of the most important responses against Echinococcus granulosus infection and, second, a procedure for the classification of patients into two groups: patients with hydatid disease and patients without a history of hydatid disease. This method was applied to 67 patients, 25 with active hydatid cysts (24 hepatic and 1 pulmonary) and 42 without a history of hydatid disease and was compared with the results obtained by conventional serology: indirect hemagglutination, latex particle agglutination, and basophil degranulation. An immunoelectrotransfer blot coupled to a discriminant analysis was more sensitive than conventional serological diagnosis and detected 100% of patients with an active hepatic hydatid cyst with a specificity of 100%. This method, however, failed to detect an uncomplicated hyaline pulmonary hydatid cyst.

Download Full-text