Prev: Calculation Lyapunov exponent from a given time series
Next: urlread and Excel-webquery example
From: Sergio Silveira on 16 Sep 2009 03:17
Hi everyone,

I'm using the System Identification toolbox (ARX and ARMAX methods, more specifically) in order to obtain regression models for my data.

My problem basically consists of an input matrix of temperatures (X) and an output vector containing natural frequencies. Their dimensions are:

X (13 x 9) - where 13 refers to the number of observations and 9 is the number of sensors (variables).
d (13 x 1) - where 13 refers to the number of observations and 1 refers to the first frequency.

So, when I use only the first column of the matrix X I do get an output model which is really fit for the ARX model, but not so much for the ARMAX. For the ARX model, I get the orders 4 4 1 and for the ARMAX, 2 2 2 1.

However, when I try to use more than two columns or the entire matrix X, I have trouble in defining the orders. When I try to use the nine columns, the orders for the ARX are predefined as such: 4 [4 4 4 4 4 4 4 4 4] [1 1 1 1 1 1 1 1 1]. In so doing, a message "There are too many parameters to estimate for this amount of data".

Any ideas?

thanks in advance.
 | 
Pages: 1
Prev: Calculation Lyapunov exponent from a given time series
Next: urlread and Excel-webquery example