The Uncertainty Propagation study type performs a number of model evaluations from LHS sampling data; such data is used to generate a surrogate model. Then a number of predictions are made by using the surrogate model in a Monte Carlo analysis. A KDE is generated with this sampled data to produce the estimation of the probability density for each QoI. For details on the KDE and surrogate method, see the Theory sections Uncertainty Propagation — Monte Carlo Analysis, Surrogate Models — Polynomial Chaos Expansion, and Surrogate Models — Gaussian Process.