This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 3.4 The bootstrap for determining confidence intervals on estimates Bootstrap allows us to go beyond the usual assumptions of linearity and Gaussianity to determine confidence on our estimates. Bootstrap is a MonteCarlo simulation technique that uses sampling theory to calculate standard errors and bias of estimators, or any other statistic we want to determine. Figure 2 describes how the technique works. First, we estimate, by MLE for example, a set of parameters θ ˆ . We know these estimated parameters are not equal to the true one, so we would like to know how good they are. Now comes the bootstrap main idea: we assume that the parameters θ ˆ are the true parameters, hence we know the true subsurface. If we also know the way errors are created from data, then we can device a Monte Carlo method to generate new “synthetic” data sets, as many as we want. These datasets are bootstrapped data (not real data). Next, we apply the same estimation technique to these bootstrapped data to get a total of B...
View
Full
Document
This note was uploaded on 01/24/2011 for the course ERE 284 taught by Professor . during the Spring '10 term at Stanford.
 Spring '10
 .

Click to edit the document details