The median and average values of importance show group aggregate, while areas and SD indicate dispersion, i.e. disagreement  of the group`s responses to this outcome. The range and SD decreased during the rounds, which showed a centralization of views, i.e. increased convergence or convergence. Comparison of meaning rankings shows similarities in median and average values, equal to or within an interval difference (Tables 4, 5, 7, 8, 9 and 10). For statements 3 and 8 (Tables 6 and 11), which were two of the statements, differences of 2 intervals were posted. These differences between means and medians could be explained by the fact that participants pay less attention to the statements they consider to be the least important. 2. Percentages for each level of agreement for each statement to compensate for different response rates, It was suggested that the consensus is the same as the agreement and that the agreement can be determined by: the evolution of the consensus was shown by the increase in percentages of agreement, convergence of the range with typical differences in meaning ratings, and a reduction in the number of comments. Stability was manifested by an upward trend in Kappa`s values. 5.
Kappa (K) weighted values to compare random matches between towers. Dajani JS, Sincoff MZ, Talley WK: stability and contract criteria for the completion of Delphi studies. Technol Forecast Soc. 1979, 13: 83-90. 10.1016/0040-1625 (79)90007-6. Chaffin WW, Talley WK: Individual Stability in Delphi Studies. Social technol prognosis. 1980, 16 : 67-73.
10.1016/0040-1625 (80)90074-8. Disc et al. suggest that stability should be used in Delphi studies to compare participants` views, as they believed that covering a percentage of the opinions expressed did not reflect the nature of Delphi in opposing a natural centralization of views. They also note that full stability will be difficult to achieve, as there will always be some “swinging movements”. A subsequent study used stability rather than consent as a termination criterion for Delphi . Dajani et al.  propose the use of Chi-quadratics (2) to test stability. However, this cannot be considered a test stability in the Delphi studies, as it will determine “the independence of the towers from the answers found in them”  and not the stability of the reactions between the different towers. Based on the work of Dajani et al. a study by Chaffin and Talley  was developed to determine individual stability before the stability of the group, and this should therefore be considered with caution. As these works appear only a few examples of 2 as a measure of stability in the Delphi literature. More recently, Greatorex-Dexter has used standard means and deviations (SD) to compare movement between Delphi towers as a measure of stability and convergence .
If necessary, statements have been reworded for the next round, with the aim of moving towards consensus on the basis of the degree of convergence and the question of the majority highlighted. Successive rounds consisted of statements (some reworded) of percentage of match levels, anonymous returns from the previous round to show the area of the views received and important rankings that use the middle rankings. Participants were invited to read the comments before responding to the statements again. Round 1 produced 5 themes. In the third cycle, all responses reflected different themes, earlier themes and new themes that indicated that participants` opinions were not exhausted. The agreement was increased over the three rounds, inversely in the important ranking. Both the median and the average define a movement towards a lesser meaning with the course of the towers. The range intervals of 6, 5 and 5 in the towers indicate that consensus was weak, but stability had arrived, but the SD was still declining, suggesting sustained convergence.