Statistical Analysis of the Hipot Experiment

For the hipot experiment, the initial ANOVA table is constructed in Table 7.15. An example is given at the top of how to calculate the sum of the squares for factor A. In order to calculate the F ratio, each variance must be compared against the error variance. Since all columns are used, and there is no repetition of the experiment, the factor with the smallest SSF is used as the source of error. This is the interaction B x D, or contact method x paint, with a sum of the squares (SSBxD) 0.79. When the F ratios are calculated for the remaining factors, not a single factor was more than 95% significant. Therefore, pooling is necessary to increase significance.

 Pooling starts with the smallest remaining sum of the squares (SSf) being added to the error SS to see if significance is achieved for

the experiment. The process is continued until no greater significance is achieved. The insignificant factors are combined with the error to obtain the pooled error. In this manner, G and A x D are pooled with error B x D. This implies that these factors, consisting of shims, and the two interactions connector x paint and contact x paint are not sig- nificant，showing that the paint operation is independent of* the rest of the factors. Only the four factors shown in Table 7.16 are significant: connector type, contact method, their interaction, and painting the box. This clearly matches the observed values in the graphical plot of factors in Figure 7.7. The factors can be ranked in importance according to the percent contribution: paint, contact method, connector, and. the interaction of connector x contact method. The total percent contribution of the error is less than 12%, indicating good confidence in the experiment. If the error percent is greater than 30%, the significance of the total DoE experiment is lessened.