The most common cause for seeing different code coverage numbers in Sandbox vs. production is creating non-portable tests.  Typically this means writing tests that assume certain records exist in an organization, for example writing a test that queries for a specific record by Id.  In Sandbox, the test passes and executes the complete code path you intended, but the same record doesn't exist in production so the test behaves differently.

The best practice here is to create all the records you need within each test method, then call System.startTest() when your actual test should start (allowing you to create test data without counting against your to-be-tested code's governor limits, though looser limits still apply).  DML statements (insert/update/delete) performed within a test method are not committed to the database, so you can do whatever you need to set up your test without worrying about impacting production data.

Also, if your triggers are not covered at all, the way to test them is to perform the appropriate DML statement within a test method.  For instance, if I want to test a before update trigger on Account, in my test method I'll want to insert and then update an Account so my trigger will fire.


You should find useful examples of test methods in the Apex Language Reference, Force.com Developer's Guide, and Force.com Cookbook.  I don't have particular chapters/examples I can point you to offhand, though.

http://wiki.apexdevnet.com/index.php/Documentation