Naslov
Application of posterior probabilities in osteometric sex estimation:a study on antique and late antique Salona population

Autori
Jerković, Ivan ; Bašić, Željana ; Kružić, Ivana ; Anđelinović, Šimun

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
6th scientific conference Methodology and Archaeometry - book of abstracts / Miloglav, Ina - Zagreb : Faculty of Humanities and Social Sciences of the University of Zagreb ; Croatian Archaeological Society, 2018, 44-45

ISBN
978-953-175-745-4

Skup
6th Scientific Conference Methodology & Archaeometry (MetArh)

Mjesto i datum
Zagreb, Hrvatska, 06.12.2018. - 07.12.2018

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Recenziran

Ključne riječi
sex estimation ; discriminant analysis ; posterior probability ; Salona ; Roman period

Sažetak
Osteometric sex estimation via linear discriminant analysis is one of the most common and the most reliable osteological approaches for sexing skeletal remains applicable also on less preserved skeletons. However, as extrapelvic skeletal dimensions are population specific, discriminant functions need to be developed for each population separately. One of the major drawbacks of this method is also a high degree of overlapping values between male and female measurements that impact sex estimation accuracy. As previous research pointed out to the application of posterior probabilities as a solution, the aim of the present study was to test that approach on the archaeological population and compare accuracy rates of discriminant functions developed by traditional and the novel approach. The sample comprised 207 skeletons from Salona necropoles dated from 1st to 6th century. We estimated sex using pelvic measurements by DSP software and took 70 postcranial measurements. We analyzed sexual dimorphism using ANOVA and selected 10% of measurements that exhibited the highest degree of sexual dimorphism. For those measurements, we developed univariate and multivariate discriminant functions and calculated accuracy using traditional approach and posterior probability threshold set at 0, 90. The accuracy of traditionally developed discriminant functions ranged from 83 to 93% while discriminant functions with posterior probability threshold reached accuracy rates ≥ 95% independent of variables that were used. Nevertheless, as the all skeletons could not meet the demanded posterior probability level, the proportion of skeletons that could not be sexed ranged from 15 to 60%. The study confirmed that application of posterior probabilities produces the highest accuracy rate acceptable not only in archaeological context but also in forensic cases. That approach also creates a certain proportion of skeletons for which sex cannot be estimated, which is why researchers, depending on research type must compromise between accuracy and proportion of sexed skeletons.

Izvorni jezik
Engleski

Znanstvena područja
Temeljne medicinske znanosti, Arheologija, Etnologija i antropologija

POVEZANOST RADA