Subscribe to RSS
DOI: 10.3766/jaaa.24.10.3
Understanding Protocol Performance: Impact of Test Performance
Publication History
Publication Date:
06 August 2020 (online)
Background: This is the second of two articles that examine the factors that determine protocol performance. The objective of these articles is to provide a general understanding of protocol performance that can be used to estimate performance, establish limits on performance, decide if a protocol is justified, and ultimately select a protocol. The first article was concerned with protocol criterion and test correlation. It demonstrated the advantages and disadvantages of different criterion when all tests had the same performance. It also examined the impact of increasing test correlation on protocol performance and the characteristics of the different criteria.
Purpose: To examine the impact on protocol performance when individual tests in a protocol have different performance. This is evaluated for different criteria and test correlations. The results of the two articles are combined and summarized.
Research Design: A mathematical model is used to calculate protocol performance for different protocol criteria and test correlations when there are small to large variations in the performance of individual tests in the protocol.
Results: The performance of the individual tests that make up a protocol has a significant impact on the performance of the protocol. As expected, the better the performance of the individual tests, the better the performance of the protocol. Many of the characteristics of the different criteria are relatively independent of the variation in the performance of the individual tests. However, increasing test variation degrades some criteria advantages and causes a new disadvantage to appear. This negative impact increases as test variation increases and as more tests are added to the protocol.
Conclusions: Best protocol performance is obtained when individual tests are uncorrelated and have the same performance. In general, the greater the variation in the performance of tests in the protocol, the more detrimental this variation is to protocol performance. Since this negative impact is increased as more tests are added to the protocol, greater test variation indicates using fewer tests in the protocol.