@LechMadeyski profile picture
Lech Madeyski @LechMadeyski

Professor at Wroclaw University of Science and Technology

Joined March 2012
  • Tweets 116
  • Following 137
  • Followers 110
16 Photos and videos
The "most accurate" LLM for systematic-review screening silently discarded 63% of the relevant papers. Our new open-access paper @ISTJrnal (Information and Software Technology) shows why standard metrics mislead — and what to use instead. 🧵 doi.org/10.1016/j.infsof.202…
1
1
93
more replies
LLM4SCREENLIT = recommendations for authors AND a one-page checklist for editors/reviewers, split by study type (benchmarking vs deployment). Validated on 9 LLMs × 24 SE secondary studies (34,528 articles). With Prof. Barbara Kitchenham & @ProfMShepperd .
1
2
21
Everything is open: 📄 Paper (CC-BY): doi.org/10.1016/j.infsof.202… 🧰 Replication package — R/Python metric scripts fillable checklist: doi.org/10.6084/m9.figshare.… Use it and adapt it.
18
Lech Madeyski @LechMadeyski
Feb 21
Personalized Share Link to our new @ISTJrnal paper "Test case prioritization: A systematic review using snowballing and TCPFramework with approach combinators": authors.elsevier.com/a/1me%7…
1
1
100
New paper in Information & Software Technology🚀 We introduce "approach combinators"—ensemble Test Case Prioritization methods improving regression testing.💡 By T. Chojnacki & @LechMadeyski 🔗 doi.org/10.1016/j.infsof.202… #SoftwareTesting #TestCasePrioritization #SoftwareEngineering

2
1
42
more replies
📊 How does it work? The graphical abstract below presents a simplified view of our framework. Test suites are passed through different combinations of simple models to produce a highly efficient test ordering—without the need for heavy computation. 👇
1
32
💡 The Results: By integrating existing strategies, approach combinators consistently improve regression testing. The ultimate takeaway? We achieve state-of-the-art TCP performance across diverse software projects! 🏆 #QA #CICD #AcademicTwitter
31
Lech Madeyski @LechMadeyski
9 Jan 2025
Exciting! e-Informatica Software Engineering Journal (e-informatyka.pl/) just published "Guidelines for Conducting Action Research Studies in Software Engineering" by @MiroslawStaron 👉🏼doi.org/10.37190/e-Inf250105 Let's make 2025 an #ActionResearch year in #SoftwareEngineering

e-Informatica Software Engineering Journal (EISEJ)

Open access journal regarding software engineering, with special focus on experimentation, machine learning and evidence based SE.

e-informatyka.pl
4
93
Lech Madeyski @LechMadeyski
9 Jan 2025
Happy to share the effect of research visit at BTH and collaboration with @wnukkris et al. We analyzed usage of requirements attributes in industry (2 industrial case studies) & literature👉🏼 doi.org/10.1007/978-3-031-70… or madeyski.e-informatyka.pl/do… #requirementsengineering #requirements

An Empirical Analysis of the Usage of Requirements Attr

Requirements attributes play an important role in storing and managing meta-information about requirements. This paper presents the results of a literature review and two industrial case studies...

link.springer.com
2
77
Interested in “Recommendations for analysing and meta-analysing small sample size software engineering experiments”? 👉🏼My joint paper doi.org/10.1007/s10664-024-1… with Prof. Kitchenham in @emsejournal #MetaAnalysis #EffectSize #NonParametric #ReproducibleResearch #SoftwareEngineering

Recommendations for analysing and meta-analysing small sample size software engineering experiments

Empirical Software Engineering - Software engineering (SE) experiments often have small sample sizes. This can result in data sets with non-normal characteristics, which poses problems as standard...

link.springer.com
1
1
5
400
more replies
Tests based on pˆ always had better or equal power than tests based on Cliff’s d, and across all but one simulation condition, pˆ Type 1 error rates were less biased.
1
54
Conclusions: Using pˆ is a low-risk option for analysing and meta-analysing data from small sample-size SE randomized experiments. Parametric methods are only preferable if you have prior knowledge of the data distribution.
49
Load more