EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Spectral Diagnostics and Sensitivity Bounds for Partial Ridge Regression in Innovation Economics

Yuang Chen, Saw Eu Gene, Kok Loang Ooi

Mathematics · 07.11.2026 · 14 (21) · 4088

DOI: 10.3390/math14214088 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mathematics, online erschienen am 07.11.2026. Verfasst von Yuang Chen und 2 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Economic analysis of technology adoption and innovation requires separating a focal technology association from correlated industry characteristics. This study develops spectral diagnostics and sensitivity bounds for partial ridge regression, with an application to artificial intelligence (AI) adoption and industry innovation. Nuisance coefficients are penalized while the focal coefficient remains unpenalized. A response-alignment bound controls changes between two penalties, a partition-based corollary controls departures throughout the interval, and a separate finite bound treats simultaneous perturbations of the response and predictors. Equality and cancellation examples delineate their scope. Gaussian experiments cover 169 settings with 5000 tuning and 10,000 independent evaluation replications per setting; six settings compare partial ridge, ordinary ridge, lasso and elastic net. Additional experiments evaluate 8400 penalty intervals and 7680 data perturbations. The economic application uses 59 United States industry cells and four innovation outcomes from the 2023 Annual Business Survey. Six controls leave 20.7% of observed AI-use variation after exact adjustment; marginal AI coefficients of 2.46–4.21 percentage points per standard deviation of AI use fall to 0.67–1.63 after full adjustment. Across 61 specifications, 3416 endpoint and uniform certificates quantify the sensitivity of these descriptive associations. A coefficient-change tolerance illustrates reporting in economically interpretable outcome units. The diagnostics assess coefficient sensitivity without identifying causal returns to AI adoption or replacing statistical uncertainty analysis.

Zitieren

APA

Chen, Y., Gene, S. E., & Ooi, K. L. (2026). Spectral Diagnostics and Sensitivity Bounds for Partial Ridge Regression in Innovation Economics. Mathematics, 14(21), 4088. https://doi.org/10.3390/math14214088

Vancouver

Chen Y, Gene SE, Ooi KL. Spectral Diagnostics and Sensitivity Bounds for Partial Ridge Regression in Innovation Economics. Mathematics. 2026;14(21):4088. doi:10.3390/math14214088

BibTeX
@article{Chen2026,
  author = {Chen, Yuang and Gene, Saw Eu and Ooi, Kok Loang},
  title = {Spectral Diagnostics and Sensitivity Bounds for Partial Ridge Regression in Innovation Economics},
  journal = {Mathematics},
  year = {2026},
  volume = {14},
  number = {21},
  pages = {4088},
  doi = {10.3390/math14214088}
}
Kostenlos starten →