EXTENSIONS TO THEORY OF TIME-SERIES ANALYSIS.
Quarterly rept. no. 9, 1 Aug-31 Oct 64,
BROWN UNIV PROVIDENCE R I
Pagination or Media Count:
During the quarter, various questions of meterological prediction have been considered. One of the difficulties of applying known techniques of statistical analysis of random processes to meteorological problems is that the processes occuring in such applications cannot be considered spatially stationary. Let us consider an idealized example. In the dynamical statistical approach one is concerned with random solutions of differential equations. The initial values of the solution ux,t were assumed to be ux,o vx. In a typical meteorological situation, the values of the meteorological quantities are observed at a sequence of weather-stations. So it is assumed that vx is observed at a sequence of points x x sub m. To solve the prediction problem, the initial field is obtained by interpolating between the values vx sub m. vx is estimated by that linear combination of the observed stochastic variables which has the smallest mean square error.