There are advantages and disadvantages to both methods.
Rendering wise: while a Polygon mesh is heavier than a NURBS mesh, it doesn't require tessellation while rendering, although more rendering time is required if you're using UV mapping.
A NURBS surface is made up of patches. During rendering, each patch is divided into an appropriate number of triangles to approximate the true shape of the surface. During render, if the mesh doesn't seem as smooth, running NURBS tessellation will add more triangles to smooth things up but will also require more rendering time.
Most of the time, NURBS are used for simpler objects, but it all comes down to using whichever you feel more comfortable with, modeling and rendering.
In the more recent version of Maya one is also able to convert from one geometry type to another to have further control while modeling on a given mesh.
For instance in the "Maya Intermediate: Photorealistic Car Modeling" training�kit one relies heavily on this feature.
Here is an example of what we do in the kit:
From your question�I would figure that you would model a car using polygons.� Using reglular poly tools such as merge vertacies, extrusions, and moving the vertacies into place, we modeled an intitial shape of the car.� Once this was finsihed, we converted the polygons to sub-d to even get further control of our mesh, adding finer detail to the car where we need it, such as the door handle, and adding the indents on the side of the Mitsubishi Eclipse. Once this was done we converted back to polygons wich was needed to render out the Car in mental ray.
I hope this helps you understand that in Maya there are many possiblities, and often the conversion of geometry types which will get you the best results, since you have different modeling tools in each object type.��Sometimes you might start with Polygons and end up with NURBS or Sub-D.