English Deutsch Français Italiano Español Português 繁體中文 Bahasa Indonesia Tiếng Việt ภาษาไทย
All categories

2 answers

A typical model is decribed here:
http://research.stlouisfed.org/wp/2003/2003-015.pdf
To model eq. 2.1 in Excel you would use the first column for t, the second for y_t. Next one column each for each component of x (which use the y column values), one column each for the draw of the random noise from each state. The last column would be used for the calculation of the state one is in (Eq. 2.2).
The y on the row k+1 uses the x-values and noise values of row k.

It is practical to leave the first ten rows or so free to define the parameters used in the calculations.

These parameters can be estimated from the data as follows.
Take out the noise colums and add a column with the measured y_t. Add another column where you calculate (y_calc - y_meas)^2.

Next, let Excel minimize the sum of this latter column, while varying your parameters. Vary one parameter at the time, as Excel is rather stupid for minimizing. Plot t, y_calc and y_meas in a scatterplot to see if you are going anywhere.

Good luck and I hope your problem is not too complicated.

2006-08-24 01:45:05 · answer #1 · answered by cordefr 7 · 0 0

What are you trying to do, estimate or simulate?

2006-08-21 13:03:17 · answer #2 · answered by NC 7 · 0 0

fedest.com, questions and answers