estimation of sex from sternal measurements in a western australian population
TRANSCRIPT
Forensic Science International 217 (2012) 230.e1–230.e5
Forensic anthropology population data
Estimation of sex from sternal measurements in a Western Australian population
Daniel Franklin a,*, Ambika Flavel a, Algis Kuliukas a, Andrea Cardini a,b,c, Murray K. Marks d,Charles Oxnard a, Paul O’Higgins a,c
a Centre for Forensic Science, The University of Western Australia, M420, 35 Stirling Highway, Crawley 6009, Western Australia, Australiab Dipartimento di Biologia, Universita di Modena e Reggio Emilia, via Campi 213, 41100 Modena, Italyc Centre for Anatomical and Human Sciences, The Hull York Medical School, Heslington, York Y010 5DD, United Kingdomd Department of Pathology, Graduate School of Medicine, The University of Tennessee, 1924 Alcoa Highway, Knoxville, TN 37920, USA
A R T I C L E I N F O
Article history:
Received 29 July 2011
Received in revised form 6 October 2011
Accepted 6 November 2011
Available online 10 December 2011
Keywords:
Sex discrimination
Forensic anthropology
Sternal measurements
Computed tomography
3D landmarks
Population standards
Sexual dimorphism
A B S T R A C T
In Australia, particularly Western Australia, there is a relative paucity of contemporary population-
specific morphometric standards for the estimation of sex from unknown skeletal remains. This is largely
a historical artefact from lacking, or poorly documented, repositories of human skeletons available for
study. However, medical scans, e.g. MSCT (multislice spiral computed tomography) are an ingenious and
practical alternative source for contemporary data. To that end, this study is a comprehensive analysis of
sternal sexual dimorphism in a sample of modern Western Australian (WA) individuals with a main
purpose to develop a series of statistically robust standards for the estimation of sex.
The sample comprises thoracic MSCT scans, with a mean of 0.9 millimeter (mm) slice thickness, on
187 non-pathological sterna. Following 3D volume rendering, 10 anatomical landmarks were acquired
using OsiriX1 (version 3.9) and a total of 8 inter landmark linear measurements were calculated using
Morph Db (an in-house developed database application). Measurements were analyzed using basic
descriptive statistics and discriminant function analyses, with statistical analyses performed using SPSS
19.0.
All measurements are sexually dimorphic and sex differences explain 9.8–47.4% of sample variance.
The combined length of the manubrium and body, sternal body length, manubrium width, and corpus
sterni width at first sternebra contribute significantly to sex discrimination and yield the smallest sex-
biases. Cross-validated classification accuracies, i.e., univariate, stepwise and direct function, are 72.2–
84.5%, with a sex bias of less than 5%. We conclude that the sternum is a reliable element for sex
estimation among Western Australians.
� 2011 Elsevier Ireland Ltd. All rights reserved.
Contents lists available at SciVerse ScienceDirect
Forensic Science International
jou r nal h o mep age: w ww.els evier . co m/lo c ate / fo r sc i in t
1. Introduction
Standard practice in situations involving the analysis ofunknown decomposed, mutilated or skeletal remains is theformulation of an osteobiography comprised of a set of biologicalattributes to facilitate personal identification [1]. Accurate sexestimation is important because it effectively reduces the pool ofpotentially matching identities and it subsequently determines themost appropriate standard to apply for estimating other biologicalcharacteristics such as age, ancestry and stature.
With a growing appreciation for the importance of population-specific standards, the literature is replete with morphometricstandards for estimating sex from a variety of bones [2–7]. InAustralia, however, there is a glaring shortage of anthropologicallyderived morphometric standards from contemporary individuals,largely because traditional repositories of long-term documented
* Corresponding author. Tel.: +61 8 6488 1232; fax: +61 8 6488 7285.
E-mail address: [email protected] (D. Franklin).
0379-0738/$ – see front matter � 2011 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.forsciint.2011.11.008
skeletons do not exist. However, in the absence of such researchcollections, medical scans and measurements on living individualsoffer an appropriate and reliable source of contemporary popula-tion-specific data from which forensic standards are beingformulated [8–11].
Peri- and post-mortem events are factors that influence whethera complete skeleton, or individual parts, are available for examina-tion. It is critical that accurate and reliable osteometric standards areavailable for an assortment of complete and fragmented bones [12].While the skull and pelvis are the most sexually dimorphic regions ofthe human skeleton and traditionally favoured when sex isestimated (e.g. [13,14]) the sternum is also sexually dimorphic,with recent research demonstrating an overall classificationaccuracy above 80% in several populations [12,15–17].
As part of a current long-term research project, we areattempting to fortify the capabilities of forensic scientists inAustralia by developing statistically quantified standards forskeletal biology from a variety of osseous landmarks. Here, wepresent whether the adult sternum can be used to estimate sex in aWestern Australian population. Our specific aims are to quantify
Fig. 1. Anterior view of a CT-scan (0.9 mm resolution) of a sternum (female – 23
years of age) marked with the measurements used in the present study; see Table 1
for key. Note: the inferior margin of the sternal body is continuous with the xiphoid
process at the xiphisternal joint; attachment is dorsal and rarely contains the entire
thickness of the body at the inferior margin, meaning there is typically no
interference with acquiring the most inferior landmark.
Table 1Definitions of the measurements used in the present study (following McCormick et al. [24] and Jit et al. [25]); measurements are illustrated in Fig. 1.
Measurement Definition
Manubrium length (M) Direct distance, from the anterior aspect and in the midline, from jugular notch to mesoxiphoidal junction
Sternal body length (B) Direct distance, from the anterior aspect and in the midline, from manubriosternal to mesoxiphoidal junction
Combined length of manubrium
and body (CL)
Sum of the manubrium and corpus lengths [M + B]
Manubrium width (MW) Width between the left and right facets for the first costal cartilage
Corpus sterni width at first
sternebra (CSWS1)
Direct distance (anterior aspect) between left and right midpoints between facets for second and third costal cartilage
Corpus sterni width at
third sternebra (CSWS3)
Direct distance (anterior aspect) between left and right midpoints between facets for fourth and fifth costal cartilage
Sternal index (SI) Calculated as the division of M by B, multiplied by 100 [(M/B) � 100]
Sternal area (SA) Calculated by multiplying the sum of M and B with the sum of MW, CSWS1 and CSWS3 divided by three [(M + B) �(MW + CSWS1 + CSWS3)/3]
D. Franklin et al. / Forensic Science International 217 (2012) 230.e1–230.e5230.e2
the reliability and accuracy of using sternal measurements toestimate sex and to outline a series of relevant forensic standards.
2. Materials and methods
2.1. Materials
This study quantifies measurements from CT scans from 187 individuals (93
males; 94 females) currently residing in Western Australia. Randomly selected
individuals were patients during 2010 and/or 2011 presenting at various Western
Australian hospitals for clinical thorax MSCT’s. The scans were anonymous when
received by the authors with only sex and age data retained. The mean male age is
25.2 years with a range of 16–34 and the mean female age is 25.1 years with an
identical range. In each scan the individual sternebrae of the mesosternum are
fused with only trace fusion lines apparent in some individuals.
While comprising various ethnic backgrounds, the sample is primarily Caucasian
which overall is representative of a ‘typical’ Western Australian population (see [18]
for specific frequency statistics). Only individuals without obvious congenital or
acquired sternal pathology (e.g. unfused ossification centres; fractures) were
included. Ethical research approval was granted by the Human Research Ethics
Committee of the University of Western Australia (RA/4/1/4362).
2.2. Methods
2.2.1. CT data and sternal measurements
Thorax patient CT imaging was performed using MSCT on a Phillips Brilliance 64
scanner (Philips Healthcare, North Ryde, Australia) with an average slice thickness
of 0.9 mm. Following 3D volume rendering, the 3D coordinates of 10 anatomical
landmarks were acquired (by AF) using OsiriX1 (version 3.9–64 bit); with a total of
8 linear inter landmark measurements calculated using Morph Db (an in-house
developed database application). Measurements are accordingly defined and
illustrated in Table 1 and Fig. 1, respectively.
2.2.2. Statistical analyses
In order to evaluate the reliability of acquiring the sternal measurements, a
precision study was performed prior to primary data collection. The sterna of the
same four randomly selected subjects were measured on four different evaluation
days, with a minimum of one day between re-measurement to minimize the
possibility of recalling figures. Intra-observer precision was evaluated by
calculating the relative technical error of measurement (rTEM) and coefficient of
reliability (R); see [19,20] for relevant definitions and formulae. Intra-observer error
is shown in Table 2 and is within accepted standards for linear measurement data
(R > 0.75; rTEM < 5%) [19,20].
Following the calculation of normal descriptive statistics, sexual dimorphism in
the measurement data was evaluated using ANOVA. The accuracy of sex estimation
using sternal measurements was then assessed using a series of jackknife leave-
one-out cross-validated discriminant function analyses (direct, step-wise and
multiple variable). Posterior probabilities of correct group membership were
calculated to further measure the effectiveness of each discriminant function. This
statistic measures the strength of a correct classification, whose values are larger
with increasing distance from the sectioning point [21,22]. All statistical analyses
were performed using the SPSS 19.0 software program.
3. Results
3.1. Univariate comparisons
The mean, standard deviation and range of the 8 sternalmeasurements are shown in Table 3. Males are clearly significantly
larger (P < 0.001) than females in all of the variables compared,which explain 9.8–47.4% of sample variance; P-values aresignificant even when multiple testing is taken into account usinga sequential Bonferroni correction [23]. The F-statistic valuesindicate that the measurements expressing the greatest dimor-phism in this sample were sternal area, the combined length of themanubrium and body, sternal body length, and manubrium width(Table 3).
Table 2Precision of the sternal measurements used in the present study.a
Measurement Coefficient
of reliability (R)
Relative technical error
of measurement (rTEM)
M 0.973 1.282
B 0.994 0.748
MW 0.943 1.394
CSWS1 0.807 2.633
CSWS3 0.992 1.763
a R and TEM are not calculated for CL, SI and SA because they are defined by M, B,
MW, CSWS1, CSWS3 (see Table 1 for definitions).
Table 4Demarking points (in mm) for sex differentiation and cross-validated classification
accuracy.
Measurementa Demarking points Expected accuracy Sex bias
Function 1 B , < 93.92 < < 83.4% 0.9%
Function 2 CL , < 141.09 < < 83.4% 0.9%
Function 3 MW , < 53.68 < < 77.5% 4.1%
Function 4 CSWS1 , < 25.32 < < 72.2% �4.6%
a Key to measurements in Table 1.
D. Franklin et al. / Forensic Science International 217 (2012) 230.e1–230.e5 230e.3
3.2. Discriminant analyses
3.2.1. Demarking points
A series of demarking points for sex estimation were calculatedusing the mean male and female values (Table 4); only thosemeasurements that yielded an expected correct assignment by sexof at least 70% with a sex bias less than 5% are shown. Demarkingpoints are simply half way between the means of females andmales and represent the threshold above or below which speci-mens are classified respectively as males or females, with anestimated accuracy given by the jackknifed classification rate; thesex-bias is the difference in correct assignment of male and femaleindividuals in the total sample – a negative value indicates thatmore males are misclassified and vice versa.
The highly significant sex difference in sternal area (Table 3)was reflected in a relatively high degree of classification accuracy(80.7%); the associated sex-bias, however, is �10.9% and thus it isnot included. Classification accuracies for the remaining non-included variables ranged from 63.6 to 66.3%; associated sex-biasvalues were 9.3 to �16.4%.
3.2.2. Stepwise discriminant analysis
In the stepwise analysis of six measurements (sternal index andarea were removed because using variables which are a combina-tion of measurements together with those same measurementsintroduces circularity into the analysis) only two were selected:combined length of manubrium and body and corpus sterni widthat first sternebrae; the accuracy of correct sex classification was84.5% with a sex bias of �3.4% (Table 5).
3.2.3. Direct discriminant functions
Two direct discriminant analyses (functions 6 and 7) wereperformed using combinations of the single variables demonstrat-ing the highest sex classification accuracy (see Table 4); discrimi-nant equations, group centroids, sectioning points, expected cross-validated accuracies and sex biases are shown in Table 6. Forfunctions 6 and 7, expected classification accuracies are 84.5 and
Table 3Descriptive statistics and comparisons of mean sternal measurements (in mm).
Measurementa Male (n = 93) Female (n =
Mean SD Range Mean
M 49.02 6.26 34.5–72.4 45.32
B 102.94 9.94 78.7–127.2 84.89
CL 151.96 12.20 118.3–183.4 130.22
MW 57.16 5.24 46.2–69.3 50.21
CSWS1 27.24 4.12 19.4–45.1 23.41
CSWS3 33.17 5.82 23.3–47.2 29.06
SI 48.02 7.54 30.7–78.7 54.32
SA 5976 932 3950–8477 4466
NS = not significant.a Key to measurements in Table 1.* Significance at P < 0.05.
** Significance at P < 0.01*** Significance at P < 0.001
84%, with sex-biases of 0.9 and �4.5% respectively. Whenattempting to classify an unknown, discriminant scores areobtained by multiplying each variable with its unstandardised
coefficient, summing them and adding in the constant. Scoresgreater than the sectioning point are classified as male whilesmaller values classify as female.
3.2.4. Posterior probabilities
The posterior probabilities of correct group membership forfunctions 1–7 are outlined in Table 7. It is evident that between 60and 83% of male and female individuals were classified above 80%certainty for functions 1–2 and 5–7 (Table 7). Lower posteriorprobability values were evident for functions 3 and 4, albeit themajority of individuals in both sexes were classified at above 60%certainty. There were no correctly classified individuals with aposterior probability less than 40% (Table 7).
4. Discussion
From single decedents to mass disaster victims, the rapid andaccurate estimation of sex is a crucial factor in facilitatingidentification. The results of this study clearly demonstrate thatsexual dimorphism in Western Australian sterni is sufficient toallow sex estimation with a high degree of expected accuracywhere the anterior thoracic cage is intact in decomposed,mutilated or fully skeletonised human remains.
All of the variables examined expressed statistically significantsexual dimorphism. The four most dimorphic variables in thesample were sternal area, the combined length of the manubriumand body, sternal body length and manubrium width; the samedimorphic pattern was revealed in a South African population byMacaluso [12]. It appears that in considering different regions inthe sternum, the length and width of the body, and to a lesserdegree the width of the manubrium, are most dimorphic. Thisfinding is confirmed by a number of previous studies, involvingmeasurement of both dry bone and radiographs, in a variety ofsamples (e.g. [12,16,17,24]). However, those significantly dimor-phic variables for calculation of sternal index and area introduce anunacceptably high sex-bias (see below) and have a negligible effect
94) F-statistic R square P-value
SD Range
4.94 33.7–62.6 20.09 0.098 ***
10.80 56.3–109.6 141.26 0.433 ***
11.49 98.2–154.7 157.46 0.460 ***
4.73 38.9–64.9 90.76 0.329 ***
3.34 16.0–33.3 48.70 0.208 ***
4.77 20.6–44.6 27.86 0.132 ***
9.45 36.6–80.1 25.34 0.120 ***
642 3085–5980 166.80 0.474 ***
Table 5Stepwise discriminant function analysis.a
Step Variables Unstandardised
coefficient
Standardised
coefficient
Wilk’s
lambda
Structure
point
Group
centroids
Sectioning
point
Correctly
assigned
Sex bias
Function 5 1 CL 0.073 0.869 0.540 0.895 < 1.031 0.006 84.5% �3.4%
2 CSWS1 0.119 0.446 0.485 0.498 , �1.020
Constant �13.355
a Key to variables in Table 1.
Table 6Direct discriminant functions using measurement combinations providing improved accuracy over the single variable (functions 1–4) standards.
Equationa: unstandardised coefficients and constant Group centroids and sectioning point Correctly assigned Sex bias
Function 6
(B � 0.083) + (CSWS1� 0.135) + �11.219 < 1.014 < 79/93; , 79/94
0.006 [84.5%] 0.9%
, �1.003
Function 7
(CL � 0.066) + (MW � 0.073) + �13.210 < 0.975 < 76/93; , 81/94
0.005 [84.0%] �4.5%
, �0.965
a Definition of measurements in Table 1.
Table 7Percentage of posterior probability intervals of correct sex classification for
functions 1–7.
Posterior probability
intervals
Males Females
n % n %
Function 1
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 6 7.7 9 11.5
0.60–0.79 18 23.1 19 24.4
0.80–1.00 54 69.2 50 64.1
Function 2
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 2 2.6 9 11.5
0.60–0.79 21 26.9 22 28.2
0.80–1.00 55 70.5 47 60.3
Function 3
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 17 23.0 7 9.9
0.60–0.79 27 36.5 24 33.8
0.80–1.00 30 40.5 40 56.3
Function 4
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 17 26.2 15 21.45
0.60–0.79 32 49.2 40 57.1
0.80–1.00 16 24.6 15 21.45
Function 5
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 3 3.9 4 5.0
0.60–0.79 10 13.0 15 18.5
0.80–1.00 64 83.1 62 76.5
Function 6
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 5 6.3 2 2.5
0.60–0.79 14 17.7 17 21.5
0.80–1.00 60 76.0 60 76.0
Function 7
0.00–0.19 0 0 0 0
0.20–0.39 0 0 0 0
0.40–0.59 3 4.0 6 7.4
0.60–0.79 13 17.1 21 25.9
0.80–1.00 60 78.9 54 66.7
D. Franklin et al. / Forensic Science International 217 (2012) 230.e1–230.e5230.e4
on the classification accuracy despite the additional informationbeing used by combining the original set of distance measure-ments.
The expected cross-validated sex classification accuracies of theeight functions outlined here ranged between 72.2 and 84.5% (Tables4–6) with the associated sex-bias for each function all below 5%. Theupper level of our classification accuracy is comparable, albeitslightly lower in some instances, to published sternal morphometricstandards for other populations (e.g. South African, </, 90.8% [12];Turkish <86% ,81% [15]; West Indian, <85.3% ,77.5% [16]; Canadian,</, 86% [17]; North Indian, <85% ,89% [25].
While standards providing high-expected sex classificationaccuracy are obviously desirable, if the latter is associated with anunacceptably large sex-bias, then the standard has limited forensicapplicability. The variables with the largest sex-biases in the presentstudy were corpus sterni width at third sternebra (�16.3%),manubrium length (�11.0%) and sternal area (�10.9%). Macaluso[12] also found that these variables have a large sex-bias: 22.6%,13.8%, and 8% respectively. Torwalt and Hoppa [17] similarlyreported large sex bias values for manubrium length (�27.3%) andsternal area (9.4%). In the present study, however, we have onlyprovided statistics for those functions with a sex-bias less than 5%.
In considering related previous research collectively, itdemonstrates that the metrical analysis of the sternum providesan accurate method of estimating sex in a number of geographi-cally removed samples. It is important to reaffirm, however, thatthe relative expression of dimorphism in this bone (e.g. [12]), as inmany other skeletal elements (e.g. skull [3,14]; pelvis [22,26]) isindeed population-specific. This means that any attempts toclassify unknown individual(s) from a population that is biologi-cally divergent from that of the original standards reference groupwill likely result in reduced classification accuracy and/or anunacceptably high sex-bias [27]. This reinforces the importance ofcontemporary population-specific anthropological standards inforensic applications.
In further considering the robusticity of our statisticalapproach, it is worth considering that any multivariate discrimi-nant analyses must be interpreted with caution, as size measure-ments are generally highly correlated and the results might not begeneralizable to non-reference samples [28]. In the present study,however, this is unlikely to be an issue because the results of our
D. Franklin et al. / Forensic Science International 217 (2012) 230.e1–230.e5 230e.5
univariate analyses using the same measurements are congruentwith those of previous studies on different populations. The directforensic application of the univariate data is straightforward anddoes not require specific knowledge in statistics; furthermore, theclassification rates are almost as accurate in predicting sex as themore complex multivariate functions (see Tables 4–6).
It is also important to note that we have not only quantified thedegree of error associated with our sexing standards (see above),but also the accuracy and precision of the raw data (measure-ments) from which they are derived. Our rTEM values for allmeasurements ranged between 0.75 and 2.63%, which iscomparable with figures reported by Macaluso [12] of 1.12 and1.55%. The main difference between studies is a slightly highervalue for corpus sterni width at first sternebra in the present study(Table 2). Irrespective, the level of measurement accuracyachieved in this study indicates that errors in precision are bothsmall and within acceptable standards (see [19,20]) and are thusunlikely to have influenced our results and the standards weaccordingly outlined.
5. Conclusion
This research is a segment of a larger federally funded(Australian Research Council) ongoing research project that isdeveloping population specific standards for Australia in the formof an interactive Human Identification Package (HIP). We haveestablished that the sternum is a suitable skeletal element for sexestimation in a Western Australian sample. A series of new forensicstandards for the estimation of sex in that population wereoutlined; cross-validated expected classification accuracies rangebetween 77.2 and 84.5%, with a sex-bias less than 5%. This studyalso affirms that medical scans of living individuals afford a reliablesource of contemporary population-specific data from whichforensic skeletal standards can be developed.
Acknowledgements
The authors would like to thank A/Prof. Rob Hart, FrontierMedical Imaging International, Western Australia, for assistancewith obtaining the CT-scans. We also offer our thanks to Dr. LenFreedman for his helpful comments on this manuscript. Funding isprovided by ARC Discovery Grant (DP1092538).
References
[1] D.A. Komar, J.E. Buikstra, Forensic Anthropology: Contemporary Theory andPractice, Oxford University Press, New York, 2008.
[2] E.F. Kranioti, M.Y. Iscan, M. Michalodimitrakis, Craniometric analysis of themodern Cretan population, Forensic Sci. Int. 180 (2008), 110.e111–110.e115.
[3] D. Franklin, L. Freedman, N. Milne, Sexual dimorphism and discriminant functionsexing in indigenous South African crania, HOMO – J. Comp. Hum. Biol. 55 (2005)213–228.
[4] G.R. Dabbs, P.H. Moore-Jansen, A method for estimating sex using metric analysisof the scapula, J. Forensic Sci. 55 (2010) 149–152.
[5] M. Steyn, M.Y. Iscan, Metric sex determination from the pelvis in modern Greeks,Forensic Sci. Int. 179 (2008), 86.e81–86.e86.
[6] V. Alunni-Perret, P. Staccini, G. Quatrehomme, Sex determination from the distalpart of the femur in a French contemporary population, Forensic Sci. Int. 175(2008) 113–117.
[7] D.T. Case, A.H. Ross, Sex determination from hand and foot bone lengths, J.Forensic Sci. 52 (2007) 264–270.
[8] R.B. Bassed, C. Briggs, O.H. Drummer, Age estimation using CT imaging of thethird molar tooth, the medial clavicular epiphysis, and the spheno-occipitalsynchondrosis: a multifactorial approach, Forensic Sci. Int. 212 (2011),273.e1–273.e5.
[9] R. Bassed, O. Drummer, C. Briggs, A. Valenzuela, Age estimation and the medialclavicular epiphysis: analysis of the age of majority in an Australian populationusing computed tomography, Forensic Sci. Med. Pathol. 7 (2011) 148–154.
[10] F. Ramsthaler, M. Kettner, A. Gehl, M.A. Verhoff, Digital forensic osteology:morphological sexing of skeletal remains using volume-rendered cranial CTscans, Forensic Sci. Int. 195 (2010) 148–152.
[11] E.F. Kranioti, N. Vorniotakis, C. Galiatsou, M.Y. Is can, M. Michalodimitrakis, Sexidentification and software development using digital femoral head radiographs,Forensic Sci. Int. 189 (2009), 113.e111–113.e117.
[12] P.J. Macaluso, The efficacy of sternal measurements for sex estimation in SouthAfrican blacks, Forensic Sci. Int. 202 (2010), 111.e111–111.e117.
[13] D. Ferembach, I. Schwidetzky, M. Stloukal, Recommendations for age and sexdiagnoses of skeletons, J. Hum. Evol. 9 (1980) 517–549.
[14] M.K. Spradley, R.L. Jantz, Sex estimation in forensic anthropology: skull versuspostcranial elements, J. Forensic Sci. 56 (2011) 289–296.
[15] U. Ramadan, S.N. Turkmen, N. ADolgun, D. Gokharman, R.G. Menezes, M. Kacar, U.Kos ar, Sex determination from measurements of the sternum and fourth rib usingmultislice computed tomography of the chest, Forensic Sci. Int. 197 (2010),120.e121–120.e125.
[16] S.A. Hunnargi, R.G. Menezes, T. Kanchan, S.W. Lobo, V.S. Binu, H.R.S. SUysal, P.Kumar, N.G. Baral, R.K. Herekar, Garg, Sexual dimorphism of the human sternumin a Maharashtrian population of India: a morphometric analysis, Leg. Med. 10(2008) 6–10.
[17] C.R.M.M. Torwalt, R.D. Hoppa, A test of sex determination from measurements ofchest radiographs, J. Forensic Sci. 50 (2005) 785–790.
[18] Australian Bureau of Statistics, Census Data, [Internet], 2006 (cited 2011 June 20).Available from: http://www.abs.gov.au/websitedbs/censushome.nsf/home/Census.
[19] R.E. Ward, P.L. Jamison, Measurement precision and reliability in craniofacialanthropometry: implications and suggestions, J. Craniofac. Genet. Dev. Biol. 11(1991) 156–164.
[20] S.M. Weinberg, N.M. Scott, K. Neiswanger, M.L. Marazita, Intraobserver errorassociated with measurements of the hand, Am. J. Hum. Biol. 17 (2005) 368–371.
[21] M.Y. Is can, M. Steyn, Craniometric determination of population affinity in SouthAfricans, Int. J. Legal Med. 112 (1999) 91–97.
[22] M.L. Patriquin, M. Steyn, S.R. Loth, Metric analysis of sex differences in SouthAfrican black and white pelves, Forensic Sci. Int. 147 (2005) 119–127.
[23] S. Holm, A simple sequentially rejective multiple test procedure, Scand. J. Stat. 6(1979) 65–70.
[24] W.F. McCormick, J.H. Stewart, J. Harlan, L.A. Langford, Sex determination fromchest plate roentgenograms, Am. J. Phys. Anthropol. 68 (1985) 173–195.
[25] I. Jit, V. Jhingan, M. Kulkami, Sexing the human sternum, Am. J. Phys. Anthropol.53 (1980) 217–224.
[26] P.L. Walker, Greater sciatic notch morphology: sex, age, and population differ-ences, Am. J. Phys. Anthropol. 127 (2005) 385–391.
[27] D. Franklin, A. Flavel, A. Kuliukas, C.E. Oxnard, M.K. Marks, Cranial sexualdimorphism and anthropological standards: preliminary investigations in aWestern Australian population, 64th Annual Meeting of the American Academyof Forensic Sciences, Atlanta, Georgia, February 20, 2012.
[28] K. Kovarovic, L.C. Aiello, A. Cardini, C.A. Lockwood, Discriminant function analysesin archaeology: are classification rates too good to be true? J. Arch. Sci 38 (2011)3006–3018.