Test Reliability When Error Scores Consist of Independent and Non-Independent Components