I carried out a comparison of the behaviours of the old and new MDA methods for obtaining 99% confidence. The later method takes account of joint probability. The initial results that I got seemed to be backwards but perhaps could be explained by increasing numbers of routers in the Internet. I went on to write a scamper driver to carry out the measurements close in time. I then carried out a run using the new driver and analysed the results. The difference in counts was small but still backwards from that expected. The difference for TCP was very small and may have been insignificant. I carried out more checking for errors in my analysis, and checked that scamper was selecting the correct mode when requested. Further checking will be carried out along the lines of counting the numbers of packets received at each router for the detected level of k, the number of successors.

The scamper analysis programs were updated to reflect the desired statistics documented recently. The statistic for UDP LBs not TCP LBs adjusted for clumps was implemented. If this proves useful the TCP version will also be implemented. All of these statistics have been adjusted to prevent counting of the same LBs more than once.