|
主讲人 |
Dag Tjøstheim |
简介 |
<p class="MsoNormal" align="left"><span lang="EN-US" style="font-size: 12pt; font-family: 'Times New Roman', serif;">The Gaussian distribution, univariate and multivariate, has a unique place in statistics. For instance, for a Gaussian distribution dependence is completely determined by correlation between pairs of variables. The idea of local Gaussian approximation is to approximate a bivariate (multivariate) density locally by a bivariate (multivariate) Gaussian density. The correlation function of the approximating Gaussian is taken as the local Gaussian correlation at this location. It is a very useful nonlinear dependence measure. In the talk I will review some recent developments, including the local Gaussian autocorrelation and its use in measuring nonlinear serial dependence in time series. Moreover, I will cover semiparametric multivariate density estimation with increasing dimension, conditional density estimation, local partial correlation and perspectives emanating from this. Illustrations will be given from recent publications and preprints.<o:p></o:p></span></p> |
