Uncertainty
This is the width of the probability distribution that includes the 95% confidence interval.  In other words, it is equal to 2 standard deviations, assuming a bell curve distribution, of the possible test results around the actual true value.

Test Uncertainty Calculation
Test uncertainty is calculated as the root sum square of the individual measurement uncertainties.  Test uncertainty = sqrt (sum of (each measurement uncertainty * each sensitivity)^2), in other words, the root sum square (RSS) or the hypotenuse of a multidimensional triangle.

Individual Measurement Uncertainties
Measurement uncertainties include the instrument accuracy, repeatability, and the spatial variation.  For example, suppose you put a perfect RTD in the gas turbine inlet.  
Since the variation from one location in the inlet and another can differ by 3F, we might estimate your uncertainty as +/-3F.  But if you install 9 RTDs, then the uncertainty may be judged to be 0.5F.

Measurement Sensitivities
Measurement sensitivities are the slopes of the correction curves.  They can also be found by calculating corrected heat rate, for example.  Then change power by 1%.  Heat rate changes 1%.  So the sensitivity is 1%/1%.

GOAL
With good selection of instruments, you should be able to get power within +/-0.4% and heat rate within 0.8%.  The uncertainty analysis will tell you the quality of the test and will tell you which instruments are most important, and where you should spend your instrument and calibration dollars.