Estimating Total Test Reliability from Parts of Unequal Length