I think if the clamp is elastic or spring tension the extra pressure would cause more wear.  If you used a fixed clamp you could manually tighten it as the wear becomes an issue.

Sorry, I meant that with a clamped-on split nut, the assembly would still wear, but would not immediately get loose as wear progressed.  Instead, you'd get near-zero backlash for the life of the part, and then replace it as a consumable item at some point down the road, hopefully a fairly long ways off.

For improving wear, how about either cutting a slit into the perimeter of the nut or completely cutting it in half, and then using some sort of external clamp to compress it onto the rod?  Then, as it wears the clamp force would keep it from getting loose.

I think if the clamp is elastic or spring tension the extra pressure would cause more wear.  If you used a fixed clamp you could manually tighten it as the wear becomes an issue.

A simple way to do this is take the threaded rod, heat it with a torch, and press it or thread it into nylon.
The heated rod will make a very nice snug imprint of the threads into the nylon which very neatly reduces backlash.

Course this does not wear as well as a proper ball screw nut.
This is a trick I've used on larger CNC machines.

Excellent idea!  I'll have to try that.

For improving wear, how about either cutting a slit into the perimeter of the nut or completely cutting it in half, and then using some sort of external clamp to compress it onto the rod?  Then, as it wears the clamp force would keep it from getting loose.

Course this does not wear as well as a proper ball screw nut.
This is a trick I've used on larger CNC machines.

I am interested to add some sort of anti-backlash to my x-axis screw on my new Solidoodle 3.  Is the screw on my machine 5/16x18 or is it 8mm?

The nut and rod are 5/16x18.  It ended up being very good at removing backlash, as I could manually move the bed up and down and feel the play before but seems rock solid with the nylon insert.  The laser pointer rocker still shows a slight amount of backlash, but much smaller than before.  Still, it wasn't enough to remove wobble on its own and I still needed firmware wobble compensation.

I'm toying with the idea of printing a 1" long nylon nut with an integrated mounting flange.  I'll post how that works after I try it out.

While I couldn't find an all-nylon nut of the right size, I purchased a 5/16-18 locknut with a nylon insert at the hardware store for 33 cents, and used a thread tap to loosen the fit slightly so it would not bind the motor but would still fit snugly.  It took a little work to open the nuthugger and replace the existing nut there, but was straightforward.

With the new nut, the bed has no noticeable play. I still have to put it through its paces, but I believe it will compare favorably to existing anti-backlash methods, which I've found tend to still allow a little play in the Z-axis, particularly if a lateral force applied to the z-rod (i.e. wobble).  I want to see how well it will wear over time, and if I can measure any backlash with my rocking laser mount (http://www.thingiverse.com/thing:53958).