Halt-Normal Plots for Multi-Level Factorial Experiments
Abstract
Half-normal plots for the interpretation of 2P factorial experiments have been developed and popularized largely through the work of Cuthbert Daniel (see Daniel 1956 and 959. In this method the 2P - 1 main effects and interactions are estimated from observations on the 2P treatment combinations. The empirical cumulative distribution of these estimates is then graphically compared with a cumulative distribution derived from a normal population. A rationale for this procedure is found in the approximate normality of the null distribution of the estimates, based upon normality of experimental errors or upon the tendency embodied in the Central Limit Theorem. According to Daniel, the half-normal plot permits the analyst to judge the reality of the largest main effects and interactions and serves to indicate bad values, heteroscedasticity, dependence of variance on mean and some types of defective randomization. The object of the present paper is to indicate and illustrate possible applications of half-normal plots to balance multi-level factorial experiments in general.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1963
- Accession Number
- ADP014617
Entities
People
- S. A. Kane