We read the Statistics Spotlight column "Fallacies of Statistical Significance" with interest. We were pleased to note that what was espoused was a common-sense approach to significance tests and data analysis. We agreed with the overall comment that, "Essentially, practical significance depends on real-world considerations..." We also seconded the recommendation to plot the data and the confidence interval.If you haven't read the column, it compares median glass tensile strength from a new furnace and an existing one. Two scenarios are considered in which glass tensile strength data are used to conduct significance tests. For each scenario, a table is constructed that shows the resulting probability of finding a statistically significant difference in median tensile strength for five sample sizes and three test significance levels. The two scenarios are:
展开▼