EViews.com

Thanks for your help.

Without wanting to state the obvious, the best way rectify your problem is to use user-supplied starting values rather than the EViews supplied ones. The options page of the Estimation dialog lets you choose user-supplied starting values, and you can set the starting values by entering values into the C vector before estimation.

By the way, you should probably update your copy of EViews, yours is quite out of date:
http://eviews.com/download/download.html

Thanks for your help.

If you can't think of values that make economic sense (you should be able to have a rough guess at sign and magnitude) you can try different sets of random numbers. In well behaved models, it should make no difference.