Matlab Database > Miscellaneous > Data Analysis > EOCtool

Matlab File(s)

Title: EOCtool
Author: Willy Doerfler
Institution: University of Karlsruhe
Description: Call:

NN ... Nx1; Number of unknowns (or the like)
ERR ... NxM; Matrix of errors
gm ... Gliding mean (fits a subset of 'gm' datapoints)
ordfac ... To scale the 'ord' result (see below)
legstr ... CELL; Legend in graphical output
outfile ... CHAR; Store the TEX-output in this file

Error constants as in (*) below in a formatted table

Graphical Output:
NN vs ERR(:,k) in 'loglog' plot

Analysis of convergence history. 'NN' is assumed to be the number
characterising the effort (eg, no of unknowns), 'ERR' a list of
errors of different types (eg, ERR(:,1) the exact, ERR(:,2) the
estimated error). 'eoctool' now computes the most probabel law
of the form:
(*) ERR(:,k) = C_{k,l} NN^{-ord_{k,l}}
for some set 'l'. 'ord' is usually called experimental oder of
convergence (EOC). Often one likes to present dependence on
'h:=NN^{1/spacedim}'. In this case use 'ordfac=spacedim'.
Keywords: convergence
File Name:
File Size: 6 KB
File Version: 1.1
Matlab Version: 6.1
Date: 2002-11-14
Downloads: 1122
Download File

Upload your own files today! Click here for more information.