While the differences in consecutive CW, or CCW, datasets seem negligible, the same cannot be said of CW and CCW data. In particular, the A-frame tilt data (see Fig.5) show a significant, 14-hump difference. However, rather than ascribe this effect to hysteresis in the track or inclinometers, and, because it is absent from CW-CCW data from the TMU, it is possibly a local A-frame effect, due perhaps to the strains in the bottom beam. It does mean, however, that using the A-frame data to predict YAW and ROLL may be problematic. We see in the next section how the prediction of ROLL from the A-frame data is complicated and how any result is indeed rather unlike that measured directly from the TMU.