Pool estimation

[andrea.neumann]

Hi,
I do have a pool data set with 109 countries and 58 years. I want to have fixed effects in time and cross section. Here come my questions:

1. Why do I get the message "near singular matrix" when using a lagged dependent as well?

2. How do I interpret the negative AKaike criterion in a loglinear model compared to the positive ones in other models?

3. I ought to do an LSDV model. What does this mean? How do I enter this in the menu for pooled estimations? What do I have to take care of? And how can I interpret such models and compare them to "normal" models?

Thank you very much.