Description of the program: lzo-test


This program makes a zeroth order ansatz and estimates the one step prediction errors of the model on a multivariate time series. This is done by searching for all neighbors of the point to be forecasted and taking as its image the average of the images of the neighbors. The given forecast errors are normalized to the standard deviations of each component. In addition to using a multicomponent time series a temporal embedding is possible. Thus one has to give two dimensions. The first is the number of components, the second the temporal embedding. This is realized by giving two numbers to the option m seperated by a comma.

Usage of the -c and -m flags

By default the first m columns of a file are used. This behaviour can by modified by means of the -c flag. It takes a series of numbers separated by commas. The numbers represent the colomns. For instance -m 3,1 -c 3,5,7 means, use three components with no additional temporal embedding, reading columns 3, 5 and 7. It is not necessary to give the full number of components to the -c flag. If numbers are missing, the string is filled up starting with the smallest number, larger than the largest given. For instance, -m 3,1 -c 3 would result in reading columns 3, 4 and 5.

Usage:

lzo-test [Options]

Everything not being a valid option will be interpreted as a potential datafile name. Given no datafile at all, means read stdin. Also - means stdin

Possible options are:

Option Description Default
-l# number of points to use whole file
-x# number of lines to be ignored 0
-m# number of components of the time series,embedding dimension 1,2
-c# columns to be read 1,2,...,'number of components'
-d# delay for the embedding 1
-n# for how many points should the error be calculated all
-S# temporal distance between the reference points 1
-k# minimal numbers of neighbors for the fit 30
-r# neighborhood size to start with (data interval)/1000
-f# factor to increase the neighborhood size
if not enough neighbors were found
1.2
-s# steps to be forecasted xn+steps=av(xi+steps) 1
-C# width of causality window steps to be forecasted
-o[#] output file name without file name: 'datafile'.zer
(or stdin.zer if stdin was read)
If no -o is given stdout is used
-V# verbosity level
  0: only panic messages
  1: add input/output messages
  2: write also the individual forecast errors for the reference points. if steps (-s)is larger than one only the largest one is printed.
1
-h show these options none


Description of the Output

The output consists of s lines, each of which containing the steps forecasted (first column) and the relative forecast errors (next columns) for each component of the vector seperately. Relative means that the forecast error is devided by the standard deviation of the vector component.

If the Verbosity level is larger than 1, the output also contains the individual forecast error for each component of each reference point. If s is larger than 1, the individual forecast errors are only given for the largest value of s.


View the C-sources.

Table of Contents * TISEAN home