Optimizing statistical evaluation of multiclass classification in diagnostic radiology: a study of the two-parameter multidimensional nominal response model

View article
PeerJ Computer Science

Main article text

 

Introduction

Materials and Methods

Dataset

Conventional NRM and 1PL-NRM

  • Pr(rij=s) represents the probability that the response of test-taker j to item i is class s,

  • The number of classes is c,

  • αis and βis are the two parameters of item i on class s (discrimination and easiness parameters),

  • θj represents the ability parameter of test-taker j.

Original MDNRM

  • s represents the ground truth of case i,

  • θjst represents the ability parameter of test-taker j in class t when the ground truth of the item is s.

2PL-MDNRM

Experiments

  • For all ability ( θ) and easiness ( β) parameters, normal distribution with mean = 0 and standard deviation = 2 was used.

  • For discrimination parameters (α), gamma distribution (alpha = 2 and beta = 2) or truncated normal distribution (a half-normal distribution with scale = 10) were used.

Evaluation

Results

Discussion

Conclusions

Additional Information and Declarations

Competing Interests

Eiji Ota is employed by Futaba Numerical Technologies. The authors declare that they have no other competing interests.

Author Contributions

Mizuho Nishio conceived and designed the experiments, performed the experiments, analyzed the data, performed the computation work, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Eiji Ota performed the experiments, analyzed the data, performed the computation work, authored or reviewed drafts of the article, and approved the final draft.

Data Availability

The following information was supplied regarding data availability:

The code and raw data are available at GitHub and Zenodo:

- https://github.com/jurader/2PL-MDNRM.

- jurader. (2024). jurader/2PL-MDNRM: 1st release (second). Zenodo. https://doi.org/10.5281/zenodo.13751068.

The third-party data is available at: doi: https://doi.org/10.1007/S11604-022-01366-Y.

Funding

This work was supported by JSPS KAKENHI (Grant Numbers: 22K07665 and 23KK0148). This work was also supported by Cross-ministerial Strategic Innovation Promotion Program (SIP) and Construction of Integrated Health Care System (Grant Number: JPJ01242). There was no role of the funding source in the study design, the collection, and analysis and interpretation of data. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

321 Visitors 304 Views 14 Downloads

Your institution may have Open Access funds available for qualifying authors. See if you qualify

Publish for free

Comment on Articles or Preprints and we'll waive your author fee
Learn more

Five new journals in Chemistry

Free to publish • Peer-reviewed • From PeerJ
Find out more