Stub areas are used when you have an area of your network that is not used for transit traffic. Traffic on a stub network always have either a source or destination address belonging to that network. With that in mind, you should be able to decide where stub areas make sense.
For example, you could use a stub area in a core distribution. Each distribution router could be placed in its own area and the core would be area 0. Since a distribution would be a stub to the core, you would do the OSPF stub stuff from the core to the dist.
A stub area is an area into which AS External LSAs are not flooded, but Type 3 LSAs are still allowed. Cisco has a feature called Totally Stubby Area or TSA. Configuring TSA prevents all routes except a default from coming into the area.
NSSA is used when you have an area that has routes that are redistributed into the area.
Most of this information came from Routing TCP/IP Vol. 1--this is a great reference.
In short, I try to take the view when designing routing is to ask what is the absolute minimum I need in the routing table to route packets. In the case of a stub network, you only need a default. If that is all you need, then design your network to accomodate that--regardless of the routing protocol.
HTH.
