Scholarlink

<p>This study evaluated third molar development in a northern Indian dataset from Uttar Pradesh (UP) using Demirjian staging and compared the results with another northern Indian dataset from Rajasthan (RS) and with the BioAlder model. The inter-rater reliability in the UP dataset was assessed using linear weighted Cohen’s kappa and showed strong agreement. Development patterns were compared conditional on age to avoid age-mimicry bias: UP versus RS was assessed using age-stratified Fisher’s exact tests and an overall ordinal-regression likelihood ratio test, and UP versus BioAlder was assessed using G-tests across all ages and within age groups. The UP versus RS datasets differed significantly overall, but age-specific differences were confined to isolated ages (12 years in males and 13 years in females). In contrast, UP diverged substantially from BioAlder, including higher stage H probabilities around age 18. Using Bayesian inversion with identical uniform age priors, the UP model yielded two- to threefold higher probabilities of being below key forensic thresholds than BioAlder: P(Age &lt; 16|G) and P(Age &lt; 18|H). These findings indicate that direct application of BioAlder to this Indian dataset would systematically underestimate the probability of being underage, increasing the risk of classifying minors as adults. For stage H, posterior probabilities were highly sensitive to the prior’s upper age bound; in our analysis, the upper bound had a greater impact on the resulting probabilities than the difference between the reference datasets.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparing an Indian dataset with BioAlder using Demirjian’s staging of the third molar

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Øyvind Bleka,
  • Parul Khare,
  • Sigrid I. Kvaal,
  • M. Siddharth,
  • Pooja Rastogi,
  • Simen E. Kopperud

摘要

This study evaluated third molar development in a northern Indian dataset from Uttar Pradesh (UP) using Demirjian staging and compared the results with another northern Indian dataset from Rajasthan (RS) and with the BioAlder model. The inter-rater reliability in the UP dataset was assessed using linear weighted Cohen’s kappa and showed strong agreement. Development patterns were compared conditional on age to avoid age-mimicry bias: UP versus RS was assessed using age-stratified Fisher’s exact tests and an overall ordinal-regression likelihood ratio test, and UP versus BioAlder was assessed using G-tests across all ages and within age groups. The UP versus RS datasets differed significantly overall, but age-specific differences were confined to isolated ages (12 years in males and 13 years in females). In contrast, UP diverged substantially from BioAlder, including higher stage H probabilities around age 18. Using Bayesian inversion with identical uniform age priors, the UP model yielded two- to threefold higher probabilities of being below key forensic thresholds than BioAlder: P(Age < 16|G) and P(Age < 18|H). These findings indicate that direct application of BioAlder to this Indian dataset would systematically underestimate the probability of being underage, increasing the risk of classifying minors as adults. For stage H, posterior probabilities were highly sensitive to the prior’s upper age bound; in our analysis, the upper bound had a greater impact on the resulting probabilities than the difference between the reference datasets.