English

Processing of Test Matrices with Guessing Correction

Machine Learning 2007-05-23 v1

Abstract

It is suggested to insert into test matrix 1s for correct responses, 0s for response refusals, and negative corrective elements for incorrect responses. With the classical test theory approach test scores of examinees and items are calculated traditionally as sums of matrix elements, organized in rows and columns. Correlation coefficients are estimated using correction coefficients. In item response theory approach examinee and item logits are estimated using maximum likelihood method and probabilities of all matrix elements.

Keywords

Cite

@article{arxiv.cs/0601087,
  title  = {Processing of Test Matrices with Guessing Correction},
  author = {Kromer Victor},
  journal= {arXiv preprint arXiv:cs/0601087},
  year   = {2007}
}

Comments

10 pages, in Russian