MAXIMUM REASONABLE ESTIMATE VARIANCE-COVARIANCE MATRICES

Polyanskiy I.S, Patronov D.Y

238 KB

The work formed the analytic expression that defines the maximum likelihood estimate of variance-covariance matrix of the observation vector of random variables distributed according to multivariate normal distribution. The solution is based on obtaining an expression that identifies the extremum point formed on the basis of the distribution of Wishart likelihood function. The obtained estimate variance-covariance matrix to enhance the evaluation, in terms of reducing the spread of error variance. That substantially increases the accuracy of the estimate variance-covariance matrix (on the order of several thousand) already on the small sample size, regardless of the bit variance-covariance matrix (size of the vector of random variables). Moreover, the computational cost estimates formed better than the existing ones. The results of the experimental verification of the analytical expressions generated maximum likelihood estimation variance-covariance matrix.

Библиографическая ссылка

Полянский И.С, Патронов Д.Ю МАКСИМАЛЬНО ПРАВДОПОДОБНАЯ ОЦЕНКА ДИСПЕРСИОННО-КОВАРИАЦИОННОЙ МАТРИЦЫ // Научное обозрение. Физико-математические науки . 2020. № 1. С. 48-49;
URL: https://physics-mathematics.ru/en/article/view?id=65 (дата обращения: 24.06.2026).