前言
创作开始时间:2021年6月18日14:47:58
更新:2022年5月25日19:56:34
这里研究一下软工中的有效性威胁。
Validity定义
有效性validity的定义:
Validity is concerned with the relationship between the experimental conclusions and reality, and how the conclusions might not be actually correct
是关于实验结果和真实之间的关系,以及结论如何可能不正确。
Validity Threats的类别
总共可分为四类:
1)conclusion validity
Conclusion validity is to do with how much the experiment setting influences the observed result. In other words, is the treatment really related to the outcome?
2)internal validity
Internal validity is about the causal relationship between the treatment and the observed result, and how confident we are that the treatment causes the result.
We first discuss the threats to internal validity, which refer to experiment bias and errors.
3)construct validity
Construct validity focuses on the relationship between the theory underlying the experiment and the outcome: how well do the theoretical approach and the result agree?
Finally, we discuss the threats to constructive validity, which refer to the suitability of the eval- uation metrics used in our study.
使用的metric是否能合理地反映出结果?
4)external validity
External validity is about how generalizable the results are outside the scope of the study, even if a causal relation is identified.
Second, we discuss threats to external
validity, which are about the generalisability of our approach.
泛化能力。
小结
创作结束时间:2022年5月25日20:02:53
参考文献
参考的文献:
- Mitigating threats to validity in empirical software engineering: A traceability case study
- Validity Threats in Empirical Software Engineering Research - An Initial Survey
- https://www.verywellmind.com/internal-and-external-validity-4584479
- https://web.pdx.edu/~stipakb/download/PA555/ResearchDesign.html