ARMAsel_mis_irreg Toolbox for use with Matlab
=============================================


Title to ownership
------------------

Any part of the ARMAsel_mis_irreg software package may freely be used for 
scientific or educational purposes. For commercial applications, 
permission is required from P.M.T. Broersen (address stated below).

How to Install
--------------

Unzip the file ARMAsel_mis_irreg.ZIP to a directory where
you want it to reside. 

You will need the ARMASA toolbox of Piet Broersen
and several files of the Automatic Spectal Analysis toolbox 
of Stijn de Waele. Both are available at
http://www.mathworks.com/matlabcentral/fileexchange
in ...  signal processing
   ...  spectral analysis

ARMAsel_mis_irreg uses the commercial MATLAB OPTIM toolbox for optimization.


Relation with the missing and irregular data software of Stijn de Waele
-----------------------------------------------------------------------

The software of Stijn de Waele is the basis of ARMAsel_mis_irreg.
The candidate model types MA and ARMA have been added.
A new algorithm for AR likelihood computation has been added that gives 
the same results for AR models, but the new software can sometimes be faster.

Multi shift slotted nearest neighbor resampling is introduced to replace
an irregularly sampled signal by several equidistant signals. The slot width 
is made smaller than the resampling distance to diminish the bias caused by
shifting the irregular observation times to an equidistant grid. 


User support
------------

Book with background of spectral analysis :

   Piet M.T. Broersen
   Automatic Autocorrelation and Spectral Analysis
   Springer-Verlag,London, 2006.
   ISBN 1-84628-328.

Journal paper about ARMAsel :

   P. M. T. Broersen, Automatic Spectral Analysis with Time Series Models, 
   IEEE Transactions on Instrumentation and Measurement, 
   Vol. 51, No. 2, April 2002, pp. 211-216.

Journal papers about ARMAsel_mis for missing data problems :

   Broersen P.M.T., S. de Waele and R. Bos
   Application of Autoregressive Spectral Analysis to Missing Data Problems.
   IEEE Trans. on Instrumentation and Measurement, vol. 53, no. 4, p 981-986, 2004.

   Broersen P.M.T., S. de Waele and R. Bos
   Autoregressive Spectral Analysis when Observations are Missing.
   Automatica, vol. 40, p 1495-1504, 2004. 

   Broersen, P.M.T. and R. Bos
   Time Series Analysis if Data Are Randomly Missing.
   IEEE Trans. on Instrumentation and Measurement, vol. 55, no.1, p 79-84, 2006.
           
   Broersen, P.M.T.
   Automatic Spectral Analysis with Missing Data
   Digital Signal Processing, vol. 16, p 754-766, 2006.

   Broersen, P.M.T.
   Who is Afraid of Missing Data in Spectral Analysis. (invited software paper)
   Preprints SYSID 2006, 14th IFAC Symp. on System Identification, Newcastle, Australia,
   p 702-707, 2006.

Journal papers about ARMAsel_irreg for irregularly sampled problems :

   Broersen, P.M.T. and R. Bos
   Estimating Time Series Models from Irregularly Spaced Data.
   IEEE Trans. on Instrumentation and Measurement, vol. 55 no. 4, p 1124-1131, 2006.

   Broersen, P.M.T. 
   Time Series Models for Spectral Analysis of Irregular Data Far Beyond the Mean Data Rate
   Measurement Science Technology, vol. 19, no 015103, p 1-14, 2008.


The program that runs by typing "simple_mis_demo" in MATLAB gives a demonstration of how 
the program can be applied most easily, with only AR candidate models and without any 
additional information. Only the parameters of the selected AR model are on the output. 

The program mis_demo has AR, MA and ARMA candidate models, between which one single model
is selected automatically. Furthermore, it demonstrates how additional information about
the models that have not been selected is obtained by adding an additional output variable.  
It gives the possibility to verify most results in the Digital Signal Processing paper. 
*** Typing :
  [asel,bsel]=ARMAsel_mis(ti,xi,ARmax); 
  psdsel=arma2psd(asel,bsel,500);
  corsel=arma2cor(asel,bsel,20);
gives 500 points of the estimated spectrum and 20 points of the autocorrelation function.
The inputs ti and xi are row vectors.
  ti are integer numbers as observation times
  xi are the observations at ti
  ARmax is the maximum AR candidate order. 

The program that runs by typing "simple_irreg_demo" in MATLAB gives a demonstration of how 
the program for irregular data can be applied most easily without any 
additional information. The program irreg_demo shows some additional possibilities.
*** Typing :
  [air,bir,sellogir]=ARMAsel_mis(ti,xi,ARmax,Tr,w); 
gives the parameters of the selected model with additional information in sellogir.
  Tr is the equidistant resampling time distance, chosen by the data analyst
  w is the slot width, as a fraction of Tr; w = 1/2 or 1/4 is often a good choice. 


Any questions on the theoretical foundations and the applicability of 
the functions can be submitted to: p.m.t.broersen@tudelft.nl

P.M.T. Broersen
Delft University of Technology
Faculty of Multi Scale Physics
Prins Bernhardlaan 6
2628 BW Delft
the Netherlands
telephone: +31 15 2786419
email: p.m.t.broersen@tudelft.nl