Find root cause of model error

[magnushelliesen]

Hi
I have a fairly big model consisting of about 15,500 equations. I try do solve it subject to exogenous values, but it stops when one variable evaluates to NA. Is there a neat way to backtrack how this NA-value came to be? Thus far I've search through the equations in text mode, but this is impractical. I'd much prefer to type in a command to show me some equation and inspect it's constituents.
Thanks!

[EViews Matt]

Hello,

A few steps can help diagnose these sorts of issues. First is identifying the observation in question. If the critical observation isn't clear from the solution series, then it's frequently easiest to run the solution with the "Display detail messages..." option enabled on the "Diagnostics" tab of the model solution dialog (or the "v=t" solution option if you're running the solution programmatically). After the solution runs, and fails, the model object's "Solution Messages" view will include an entry for every solved observation as well as the failed observation, making identification easy.

Next, a bit a manual detective work beginning with the equation generating the NA. You can inspect all variables the failing equation depends on at the appropriate observations, which should provide enough information to determine why the NA is occurring. It may be necessary to repeat this step for some of the upstream variables if the reported equation isn't the true source of the numerical issue. Much will depend on the nature and complexity of the model.