This is machine translation

Translated by Microsoft
Mouseover text to see original. Click the button below to return to the English verison of the page.

Note: This page has been translated by MathWorks. Please click here
To view all translated materals including this page, select Japan from the country navigator on the bottom of this page.


Standard errors for mean and covariance of incomplete data


[StdMean,StdCovariance] = ecmnstd(Data,Mean,Covariance,Method)



NUMSAMPLES-by-NUMSERIES matrix with NUMSAMPLES samples of a NUMSERIES-dimensional random vector. Missing values are indicated by NaNs.


NUMSERIES-by-1 column vector of maximum-likelihood parameter estimates for the mean of Data using the expectation conditional maximization (ECM) algorithm


NUMSERIES-by-NUMSERIES matrix of maximum-likelihood covariance estimates for the covariance of Data using the ECM algorithm


(Optional) Character vector indicating method of estimation for standard error calculations. The methods are:

  • hessian — (Default) Hessian of the observed negative log-likelihood function.

  • fisher — Fisher information matrix.


[StdMean, StdCovariance] = ecmnstd(Data,Mean,Covariance,Method) computes standard errors for mean and covariance of incomplete data.

StdMean is a NUMSERIES-by-1 column vector of standard errors of estimates for each element of the mean vector Mean.

StdCovariance is a NUMSERIES-by-NUMSERIES matrix of standard errors of estimates for each element of the covariance matrix Covariance.

Use this routine after estimating the mean and covariance of Data with ecmnmle. If the mean and distinct covariance elements are treated as the parameter θ in a complete-data maximum-likelihood estimation, then as the number of samples increases, θ attains asymptotic normality such that


where E[θ] is the mean and I(θ) is the Fisher information matrix.

With missing data, the Hessian H(θ) is a good approximation for the Fisher information (which can only be approximated when data is missing).

It is usually advisable to use the default Method since the resultant standard errors incorporate the increased uncertainty due to missing data. In particular, standard errors calculated with the Hessian are generally larger than standard errors calculated with the Fisher information matrix.

    Note   This routine is slow for NUMSERIES > 10 or NUMSAMPLES > 1000.

Introduced before R2006a

Was this topic helpful?