Consult the ANOVA table and MLR model diagnostics in conjunction with the use of the Half-Normal Probability plot for the final selection of significant factors for the model. So selecting the factor points which lie reasonably off of the line describing insignificant factors is an easy graphical way to identify important factors and start the process of optimizing the model. A red line through the insignificant factors helps to graphically delineate the difference between significant and insignificant factors. The points for factors with a 'large' and significant effects will visually fall off of the straight line described by the insignificant factors. The points comprising factors with small and/or insignificant effects on the response will describe (roughly) a straight line on the plot. The absolute value of a factor's effect is the value plotted on the x-axis, but the color of the data points indicates whether the original effect is positive (red) or negative (blue). A half-normal distribution is the distribution of the abs(X) with X having a normal distribution with mean zero. Thus, the y-value for the jth effect is the half-normal probability value for the jth value (rank) in a variable with N observations. They are given by the idealized expected values for this number of effects, ranked by increasing value, if they were drawn from a half-normal distribution. The y-values are not based on the DOE data. The Standardized Effect for a factor is the difference of the average response variable over "high" factor levels minus the average response over the "low" factor levels. This plot shows the magnitude of the experiment's effects as “Standardized Effects”, ordered in increasing magnitude, along the x-axis. Clicking on the 'Half-Norm' menu button will open a Half-Normal probability plot. For more information, go to Weibull distribution.The Half-Normal plots is a graphical tool used to help identify which experiment factors have significant effects on the response. The Weibull distribution is also used to model skewed process data in capability analysis. For example, the distribution is frequently used with reliability analyses to model time-to-failure data. Weibull The Weibull distribution is a versatile distribution that can be used to model a wide range of applications in engineering, medical research, quality control, finance, and climatology. For more information, go to Exponential distribution. Independent events are assumed to occur at a constant rate. Exponential Use the exponential distribution to model the time between events in a continuous Poisson process. For more information, go to Lognormal distribution. The lognormal distribution is used for reliability analysis and in financial applications, such as modeling stock behavior.
Use the lognormal distribution when random variables are greater than 0. Lognormal A random variable follows the lognormal distribution if the logarithm of the random variable is normally distributed. For more information, go to Normal distribution. Many statistical analyses assume that the data come from approximately normally distributed populations. Normal The normal distribution is the most common statistical distribution because approximate normality occurs naturally in many physical, biological, and social measurement situations. To fit a lognormal distribution, an exponential distribution, or a Weibull distribution, all data values must be greater than 0.