Conditional Expectation and Unbiased Sequential Estimation

By D. Blackwell. 1947

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms (2019, methodology): “In addition to the gradient estimate provided by Proposition 6, we can also derive a Rao-Blackwellized (Rao, 1992; Blackwell, 1947) estimate of the gradient of the regret, based on the formulation derived in Proposition 5.”
A Statistically Precise and Relatively Simple Method of Estimating the Bio-Assay with Quantal Response, Based on the Logistic Function (1953, result): “By an extension of Blackwell's theorem [9] to biased estimates, these estimates have a mean square error equal to or smaller than the original estimates, and their bias, if any, will be the same as that of the original estimates.”
Using MCMC chain outputs to efficiently estimate Bayes factors (2011, methodology): “Although the Rao–Blackwell theorem applies to unbiased estimators, one would expect that the general strategy of using all the information available to estimate a quantity would yield better estimates than any strategy that does not.”
Completeness, Similar Regions, and Unbiased Estimation—Part II (2012, unknown): “The result of Rao (1945) and Blackwell (1947) of which this theorem is a con sequence states that if T is a sufficient statistic for |lx and V in V an unbiased estimate of g(0), then if i/r is defined by i/r(f) = E(V\t), <*> = \?r(T) is also an unbiased estimate of g(0) with Var^ (^)< Var^ (V),…”
Multi-Prediction Deep Boltzmann Machines (2013, background): “Maximum likelihood should be better if the overall goal is to draw realistic samples from the model, but generalized pseudolikelihood can often be better for training a model to answer queries conditioning on sets similar to the Si used during training.”
A Distributed, Architecture-Centric Approach to Computing Accurate Recommendations from Very Large and Sparse Datasets (2011, background): “The Rao-Blackwell theorem states that if g(x ) is an estimator for θ, then conditional expectation of g(x ) given a sufficient statistic T (x ) is a better estimator of θ and never worse [25].”