I took the difference and spike data that IMC had prepared and calculated two kinds of trends. First, I chose a single parameter to represent the "spikiness" or "differentness" of the dataset. For the spikes, I used the RMS of the six spike sizes IMC had given. For the "differentness" I used the RMS of just the four terms that affect the pointing model. Then I found my trends. One is a simple running mean - for example, the mean associated with the 40th data set is the mean of datasets 1 to 40. The second is a weighted mean; for the first dataset I used its value as the mean, for subsequent datasets I used a value found from

New average = (current value + previous average*r)/(1+r)

where r was set to 10.

The first plot shows the mean differences:

And the next plot shows the spike data

The data for the last six months are shown here:

Conclusions

In both cases the weighted mean shows a downward trend in the last month or so. Apart from that, there seems to be a real difference between the way these two metrics of inclinometry change have evolved. The "difference" data (simply the rms of a point-by-point comparison of datasets) does not seem to have got much worse over the last 2 years. The spikes, however, have grown at what may be an accelerating pace broken only by the recent fallback.

Justin Greenhalgh

12/16/98