Odontometric sex assessment from logistic regression analysis

Odontometric sex assessment is considered a useful adjunct to more robust predictors such as pelvic and cranial bones, and discriminant function analysis (DA) has been widely applied in dental sex assessment. Logistic regression analysis (LRA) is considered a better alternative, although still untested in odontometric sex prediction. This study examines the use of LRA in dental sex assessment and compares its success to DA. Mesiodistal and buccolingual dimensions of all teeth, except third molars, were obtained on dental stone casts of 105 young adults (52 females, 53 males) using digital caliper. Application of LRA to teeth of both jaws combined and to maxillary and mandibular teeth separately yielded correct sex allocation rates ranging from 76% to 100%, which proved superior to sex assessment using DA (similar to 52-71%). LRA enabled optimal sex prediction (100%) when all teeth in both the jaws were included. Results were not as accurate when only maxillary (76.2%) or mandibular (84.8%) teeth were used. To assess and compare the use of these multivariate techniques in practical forensic casework, > 25% of tooth variables were randomly deleted. LRA still performed better (similar to similar to 91% sex allocation accuracy vs. 62.9% for DA), indicating that LRA may be superior in its ability to predict sex irrespective of the presence of complete or incomplete sets of dentitions and should be preferred in dental sex assessment. The 100% success rate of LRA in correctly assigning sex is also noteworthy considering that, in general, tooth measurements have yielded sub-optimal sex prediction levels. However, unambiguous sex assessment is possible only when the entire dentition is available and correct sex allocation levels decreases when teeth are missing.