Bias Resulting From The Study Of Selected Samples
Matthew Chukwuma Michael
Department of Mathematics and Statistics, Delta State Polytechnic, Ogwashi-Uku, Delta State, Nigeria
Oyeka I C A
Department of Statistics, Faculty of Physical Sciences, Nnamdi Azikiwe University, Awka, Anambra State.
Keywords: Antecedent factors, biases conditional probabilities, specific rates association, fourfold table
Abstract
This study developed and presented the method for estimating the biases resulting from the study of selected samples in a population. The algebra underlying the method is a simple manipulation of conditional probabilities in a fourfold table. The entire goal of the study is divided into two namely; to develop the method for determining the proportion, out of those who possess the condition and are hospitalized, who possess the antecedent factor and; to develop the method for determining the proportion, out of those who possess the condition and are hospitalized, who do not possess the antecedent factor. Specifically the method was illustrated with a situation where not all subjects are equally like to end up in the study sample and bias results when the association found in the selected samples are presumed to apply in the population at large. It was found that the two probabilities are not equal if the rates of hospitalization are not, even though the corresponding probabilities of persons having the antecedent factor given that they have the condition and having the antecedent factor given that they do not have the condition in the community are equal.