I would build something you care about - something in a task or problem domain you are interested in. Otherwise when the going gets tough, you say, "why bother" or "there's already something out there that does this."
If you're into stats, a good project might be to write some PHP/HTML/JS to query a dataset and run some analytics (mean, mode, filtered by date, other criteria). I'd also suggest doing some dynamic AJAX style interaction with jQuery. That way you're getting a pretty good sense of how PHP works, and how to use it to render something interesting in HTML.
Best of all, you may solve an actual problem that you or a colleague has, rather than writing throw away code with no value.