The algebraic formula for the resulting μ-control analysis problem is readily calculated and allows the computation of the amplitude of a worst case sensitivity frequency response for explicit evaluation of the prospective robust control performance.
In the worst case (sensitivity estimate of 0.33), a true OR of 1.50 would decrease to 1.32.
We conducted further analyses in order to determine the effects of maximum over and under estimates of wear to give worst outcome sensitivity and specificity values.
A "worst case" sensitivity analysis, assuming that fractures were present in the four patients who were lost to follow-up and in all patients with effusions, gave an overall sensitivity of 95.3% for the detection of fracture.
With sufficient 12-week data, the impact of missing data on the main outcomes will be assessed using best and worst case sensitivity analyses, using multiple imputation [ 21], or otherwise using the linear mixed effects model with the available 4-week and 8-week data to model the 12-week effect from the trend in the outcome over time in each arm.
Initiated at the interface of Information Theory and Robust Control about fourteen years ago, the approach employs the a-anisotropic norm of a linear system as its worst-case sensitivity to input random disturbances whose mean anisotropy is bounded by a nonnegative parameter a.
For all analyses, we assumed that any participant enrolled in the study who failed to return for follow-up continued their reported risky behavior, constituting a worst-case sensitivity analysis.
Even in the "worst-case" sensitivity analysis, the treatment effect of COX-2-selective inhibition remained significant.
Attributing the primary end point to all LTFU patients (worst-case sensitivity analysis), celecoxib remained superior (OR 1.46; 95% CI 1.18-1.82; p = 0.0006).
If the amount of missing data seems problematic (eg, over 20%), we will assess the robustness of the results by data imputation in tandem with best- and worst-case sensitivity analyses.
