Brewer, Robert F. (1992, ASQC) Industrial Design & Engineering Associates, Lakeville, PA 18438-0209
Analysis of Variance (ANOVA) techniques do not reveal the shape of distributions or indicate how these distributions relate to specifications. This paper discusses practical applications of multiple probability plots in designed experiments. These probability plots can help identify sources of variability and the importance of the shape of the distribution in analyzing experimental data.
A designed experiment, which is essentially a designated group of batch processes performed under controlled conditions, is a useful technique for manipulating and examining variables in order to determine their effects. A Probability Plot, which is plotted on special graph paper which shows a normal distribution as a straight line, illustrates several types of distribution information: estimates of data parameters, percentiles, and percentages between limits. The deviation from the normal distribution plot identifies whether the distribution: (1) is skewed positively or negatively; (2) is truncated; (3) is a multi-model distribution; or (4) contains outliers.
Because Probability Plots can compare various types of distributions on one sheet of graph paper, they are effective for analyzing designed experiments and for providing clues regarding process improvement. Another advantage of Probability Plots is that they are easy for non-statisticians to understand and, therefore, useful when presenting results.
Analysis of variance (ANOVA),Design of experiments (DOE),Graphical methods,Statistics