Comparison of the various methods used

To compare all the particle size distributions measured with the various sets, an average distribution could be calculated for each measurement method. The recommended procedure to calculate this representative curve should be to use the Bootstrap method to calculate the mean and the two sided 95 % confidence limits, given the data for each measurement method. But as only the LAS-W has enough data to justify such a procedure, the average median diameters will be used as a comparison. Also, the calculation of median values from only two sets and in some cases one set should be used with caution. In an ideal situation, all the median diameters should be identical regardless of the method used. It is clear that is not the case in Figure 5.

Figure 5: Bar graph of the median diameters for all the methods.

The following observations can be made:

• If only the LAS-D data from set V is considered, then the data corresponds very closely with the LAS-W method. As this similarity between the two methods is based on only one data set for LAS-D, more testing is needed to verify this similarity.
• The EZS does not compare with the other methods. However, it should be kept in mind that the only value that we considered was set W, which was considered an outlier for SRM 114p. The other set available was R, which was also considered an outlier for SRM 114p.
• Depending on the cement used, the SEM method median value is either higher or lower than the average value. Only once for cement CCRL 131, the SEM value is similar to the others. But again this is the result from only one set, therefore, there is not enough evidence for indication of variability. It should be also noted that the SEM seems to "see" fewer small particles and more larger one, as shown by the graphs in Appendix E. This is probably due to agglomeration of the small particles.
• The sedimentation method seems to follow the average of all methods. It should be noted that measurements below 3 µm in size were reported. As we do not have more details, it is impossible if this is a limitation due to the method or the reporting.

These are very preliminary findings because of the limited sample size (i.e., number of data) for most of the measurement methods and the lack of knowledge about the parameters used to interpret the measurements.