Cramér–Rao bound
The Cramér -Rao inequality, named after the two mathematicians Harald Cramér and Calyampudi Radhakrishna Rao, provides in mathematical statistics at any parameters to estimate a lower bound for the variance of an estimator with respect to this parameter.
Definition
Be parameter random variable with unknown density estimator. Be the Fisher information. Then the Cramér -Rao inequality holds
,
Provided that the following regularity conditions are met:
- The carrier of the probability densities is not dependent on the unknown parameters.
- The probability densities are steadily differentiable.
Efficiency and optimality
If the identity of an estimator
Applies, it is called efficient. If he also is unbiased, it is optimal with respect to the mean square deviation. For an unbiased estimator, the lower bound on the inverse Fisher information easier.
Regularity and proof idea
The proof of the Cramér -Rao inequality is mainly due to the Cauchy- Schwarz inequality and two model assumptions governing the commutativity of differentiation and integration.
On the one hand
Apply and on the other hand, we take
Of. Direct insertion into the Cauchy- Schwarz inequality then yields the assertion.
Multidimensional formulation
Under similar regularity conditions, the Cramér -Rao inequality can be formulated also in the case of multi-dimensional parameters. The statement is then transferred to the consideration of the covariance matrix of the multivariate estimator and provides a relation in the sense of Loewner - order matrices.
Be the vector of unknown parameters and a multivariate random variable with an associated probability density.
The estimator
For the parameter vector has a covariance matrix
The Cramér -Rao inequality in this case is
The Fisher information matrix
Is.
Applications
Using the Cramér -Rao inequality can be the dynamic permeability of membranes to assess what is brisk application mainly in biotechnology and nanotechnology.