Error propagation techniques

The propagation of errors through nonlinear systems using different error propagation techniques are discussed in this lecture. The Sandwich methodology of error propagation is widely used in many useful computation in the analysis of data uncertainties. However it involves the linearity assumptions. Unscented transformation, an efficient, consistent and unbiased transformation procedure suggested by Julier and Uhlmann can be used for error propagation studies. UT method is superficially similar to Monte Carlo method but uses a small deterministically chosen set of sample points which are selected according a specific deterministic algorithm. It was shown that this deterministic method of UT produces better results compared to that of sandwich formula, for nonlinear error propagation. (author)