Stature estimation from different combinations of foot measurements using linear and multiple regression analysis in...

Establishing the identity of the deceased is the most important task for forensic anthropologists in forensic case-work involving unidentified human remains. In such cases, forensic anthropologists examine the remains to derive the biological profile of the deceased i.e. estimation of age, sex, stature, and ethnicity to narrow down the search of the missing. Dismembered remains are recovered in mass disasters such as train mishaps, airplane crashes, earthquakes, and terrorists' attacks or in homicidal cases where perpetrator intentionally mutilates the dead body to conceal the identity of the victim. Stature estimation is considered as one of the most important tasks when a mutilated foot is recovered in process of narrowing down the pool of possible suspects/victims. Allometry is the underlying principle for estimation of stature from foot dimensions. It has been learnt from the published literature that multiple regression models including more than one factor enhances the estimation accuracies. Among the various foot dimensions, foot length is the most frequent parameter used in the estimation of stature in forensic literature. In the present study, an attempt has been made to standardize the stature estimation models from various possible combinations of foot dimensions. For this purpose, 388 Jatt Sikh males aged between 18 and 30 years were recruited from various villages of Ludhiana district of Punjab State in Northern India. Stature, five foot length measurements, and two foot breadth measurements were taken on each subject. Linear and multiple regression models were derived for the estimation of stature from various foot measurements. The highest coefficient of determination and estimation accuracy (the least standard error of estimation S.E.E) was observed from T1 (R² = 0.397, S.E.E = 4.7109) when a single foot dimension was included in the model, (R² = 0.416, S.E.E = 4.6425) from (T1, T3) when two-foot lengths were taken, (R² = 0.418, S.E.E = 4.6426) from (T1, T3, T4) when three-foot lengths were included, (R² = 0.418, S.E.E = 4.6473) from (T1, T3, T4, T5) when four-foot lengths were included, and (R² = 0.418, S.E.E = 4.6531) when all the five foot lengths (T1, T2, T3, T4, T5) were included in the regression model. It has been concluded that multiple regression models provide more accurate results than linear regression models. However, inclusion of a factor having a weak correlation with stature in the regression model, decreases the accuracy of the model.