Studies in Support of the Application of Statistical Theory to Design and Evaluation of Operational Tests. Annex C. A Cost Optimal Approach to Selection of Experimental Designs for Operational Testing under Conditions of Constrained Sample Size.
Abstract
The problem of selecting the specific design structure for an operational test under conditions of constrained sample size is investigated. The research is limited to univariate, quantitative, continuous, linear response models. Expected additional system cost (EASC) is defined and modeled as the sum of four general cost elements: fixed cost of testing; sampling cost, which is assumed to be linear; expected cost due to Type I error; and expected cost due to Type II error. The general form of each variable element is investigated as it is affected by the type of response model and by the general design structure. The problem of selecting the significance level, the number of observations for each treatment and their allocation among the cells within treatments with limited sample sizes is formulated as a constrained, nonlinear optimization problem with EASC as the objective function. The specific formulation for the 2 to the nth completely crossed, fixed factor design with a single covariate is developed and used to illustrate the derivation of a solution procedure based on a method of partial enumeration.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1977
- Accession Number
- ADA052075
Entities
People
- Sam Wallace Russ Jr
Organizations
- Georgia Tech