Login
Login

Re: Trouble in using ARIMASim

Home 21090308 Forums Re: Trouble in using ARIMASim

#2140
Ryu
Member

You will need to have some basic understanding of Time Series Analysis before you can use SuanShu in a productive way.

In general, there are many ways to take care of the trend and seasonality in a time series data. One way to do it is to use [tt:175b0jhv]MADecomposition[/tt:175b0jhv]. This is equivalent to the R function [tt:175b0jhv]decompose[/tt:175b0jhv]. They both decompose a time series into seasonal, trend and irregular components using moving averages.

See
http://www.numericalmethod.com/javadoc/suanshu/com/numericalmethod/suanshu/stats/timeseries/linear/univariate/stationaryprocess/MADecomposition.html
http://stat.ethz.ch/R-manual/R-devel/library/stats/html/decompose.html

For example,

Assuming you can take care of those somehow (e.g., as in the above), and assuming the ARMA model for the stationary part of the data, you can then do an estimation to find the ARMA specification. One way to do it is by conditional sum of squares.
http://www.numericalmethod.com/javadoc/suanshu/com/numericalmethod/suanshu/stats/timeseries/linear/univariate/stationaryprocess/arma/ConditionalSumOfSquares.html

Here are some examples of fitting the stationary part of the data using ARMA model.

Hope this helps.