Download as pdf or txt
Download as pdf or txt
You are on page 1of 58

TA3: Past, Present,

& Future
Sara M. Getz, PhD, D-ABFA
Assistant Professor of Forensic Investigation - University of Wisconsin-Platteville
Forensic Anthropologist - Oklahoma Office of the Chief Medical Examiner (Jan. 2023)

SOFA / FTCoE Webinar – Dec. 2022


Presentation Overview
• Development of the TA3 Concept
• Trait Selection and Refinement
• TA3 Traits & Trait Manual
• TA3 Reference Data
• Methods for Estimating Age from TA3 Data
• Ongoing & Future Work
The Beginning…
• 1996 – TA research begins
• recognition that other “low-information” traits may
be useful for age estimation à list generation
• 2002 – Boldsen et al. method published
• components of cranial sutures, pubic symphyses,
and iliac auricular surfaces
• new statistical framework
Validation Study of Boldsen et al. 2002

maximum likelihood
point estimate

identity line:
if known age = estimated age

Milner & Boldsen (2012) Milner Data


Am J Phys Anthropol documented age from Collection: Bass &
(148): 98-110, Fig. 8 collection records Mercyhurst (N=252)
Validation Study of Boldsen et al. 2002

Milner & Boldsen (2012) Milner Data


Am J Phys Anthropol Collection: Bass &
(148): 98-110, Fig. 8 Mercyhurst (N=252)
Validation Study of Boldsen et al. 2002

under-estimation

Milner & Boldsen (2012) Milner Data


Am J Phys Anthropol Collection: Bass &
(148): 98-110, Fig. 8 Mercyhurst (N=252)
documented
age

Pubic Symphysis &


estimated age interval
Iliac Auricular
max. likelihood
Surface = Limited
Age Information in
Parts of the Lifespan

Milner & Boldsen 2012 - AJPA 148 (1): 98-110


Pubic Symphysis &
Iliac Auricular
Surface = Limited
Age Information in
Parts of the Lifespan

Milner & Boldsen 2012 - AJPA 148 (1): 98-110


Issues with
Existing TA (ABDOU)
• relies only on cranial sutures, pubic
symphysis, and iliac auricular surface
• indicators often damaged or not preserved
• traits provide little information after age 50
• cranium contributes minimal (if any) data
• systematic point-estimate bias and
individual estimates are often inaccurate
• age ranges are often wide (poor precision)
Experienced Observer Using
“Low-Information” Traits

Milner & Boldsen (2012) Am J Phys Anthropol (148): 98-110, Figs. 8 & 14
My History with TA3
• 2007 – undergraduate
honors thesis research
using “new traits to
code” list

2007 • 2009 – age-related


research
2012
2009 • 2012 – doctoral work
focusing on
development and use
of TA3 traits
Example Transition (Logistic Regression)
humerus_lesser_tubercle_lippingL
Plot

1.0
● ●●● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

age-at-transition:

0.8
equal probability of

Probability (Transition)
absent or present
logistic

0.6
curve

95% CI
0.4
0.2
0.0

● ● ● ●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●● ● ●● ● ●● ●●●● ●● ●

20 40 60 80 100
Age
Trait
Refinement
Process

Getz Dissertation (2017)


Preliminary Trait Refinement:
Getz Dissertation
2014: Two TA3-Related Projects
Forensic Archaeological

• Getz Research Assistant / Postdoc • Getz Dissertation


• six observers • one observer
• multiple ancestry groups – same • one ancestry group (broadly
time period defined) – different time periods
• forensic anthropology • bioarchaeology
NIJ Grant # 2014-DN-BX-K007:
TA3 International Research Team

Peter Tarp George Milner Sara Getz Svenja Weise Jesper Boldsen Steve Ousley
Desirable Aspects of Traits
• scoring areas should be easy to locate (even in
fragments)
• trait definitions should be unambiguous
• traits should be distributed in as many parts of the
skeleton as possible
• trait age-of-transitions should collectively be
distributed throughout adulthood
NIJ Grant # 2014-DN-BX-K007:
Data Collection
Categories of Features
1) Descriptive features
- present / absent based on description

2) Features with non-metric thresholds


- number of features or a
fraction/percentage of a defined area

3) Features with metric thresholds


- form of a single measurement, two
measurements, or an area
Feature Types
• Epiphyseal Fusion / Developmental traits
• Shape
• “Lipping”
• “Bumps” / Bone Deposition on Surfaces
• “Spurs”/Spines/Crests
• Macropores
• Weights & Measures
• “familiar” traits of PS and AS
Terminology & Definitions
Epiphyseal Fusion
superior and inferior surfaces of L1 and L5
iliac crest a
medial clavicle
b

a b
Other Development
Features
spheno-occipital synchondrosis
S1-2
ventral rampart
a b
Shape
parietal depression
humerus lesser tubercle
L1 Spinous Process
L5 superior & inferior
S1 margin
Rib 2 and 3-10 – body
Rib 2 and 3-10 – end shape
sacrum
a) b) c)
a2 b2

Shape Cont.
b

a b c
Lipping
occipital condyles
C1 dens facet
glenoid fossa
radius tuberosity
trapezium
humerus medial and lateral epicondyles
femoral fovea
posterior acetabulum
acetabulum inferior joint
C, T, and L vertebrae

R R
Macropores
medial clavicle
lateral clavicle
humerus greater tubercle
inferior portion of auricular surface
Beware of
Subchondral
Defects
“Bumps” / Bone Deposition
medial clavicle
humerus lesser tubercle
humerus medial and lateral epicondyles
femoral head
greater trochanter – lateral surface
AIIS
posterior acetabulum
ischium
posterior sternal body
a

a b
Spurs / Spines / Crests
radius tuberosity lateral fibula ischium (superior
olecranon process Iliac crest and medial)
trochanteric fossa

L1 R1

L2 R2

b
Features of the Pubic Symphysis
surface morphology
superior protuberance
ventral and dorsal margins
symphyseal collar

**NOT the same as existing TA**


Reduced # of categories and revised
definitions.
Features of the Iliac Auricular Surface
3 regions of exostosis and
inferior joint porosity
**NOT the same as existing TA**
Weights & Measures
• humerus
• os coxa
• tibia
• calcaneus
b

a b

R L
TA3
Reference
Data
Data Collection Considerations
• Skeletal samples: reliable documentation, males and
females, 15+ yrs., diverse ancestry

• Data collection: blind recording, teams of 2

• Skeleton exclusion criteria: erroneous documentation,


unresolvable commingling

• Trait exclusion criteria: obvious pathology or trauma to


element being scored or directly articulating surface
clav_med_epiphysisL_01_23 scapula_glenoid_fossaL_0_123 fovea_marginL

1.0
● ●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

1.0

1.0
● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ● ●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

0.8
0.8

0.8

Probability (Transition)
Probability (Transition)

Probability (Transition)

0.6
0.6

0.6

0.4
0.4

0.4

0.2
0.2

0.2

0.0
● ● ● ●● ● ●●● ●●● ● ●●●● ●●●●●● ● ● ●●● ● ● ● ●● ●● ● ●

0.0

0.0
● ● ●●● ● ● ● ●● ● ● ●● ● ●

20 40 60 80 100
20 40 60 80 100 20 40 60 80 100
Age

Traits Selected
Age Age

medial_trochL humerus_lesser_tubercle_lippingR C1_eburnation

1.0
● ● ● ● ●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

1.0

1.0
● ● ● ● ●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

With Transitions

0.8
0.8

0.8

Probability (Transition)
Probability (Transition)

Probability (Transition)

0.6
0.6

0.6

0.4
0.4

0.4
Throughout

0.2
0.2

0.2

0.0
● ●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●● ● ● ●● ●

0.0

0.0
●●● ● ● ●●●●●●●●● ●●●●●●● ●●● ●●● ● ●●● ● ●●●● ●●●● ●●●● ●● ● ● ●●● ●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●● ● ●● ● ●● ● ●● ●

Adulthood
20 40 60 80 100 20 40 60 80 100 20 40 60 80 100

Age Age Age

sternal_dorsal_ridges trapezium_lippingL calcaneous_weight

1.0

1.0
● ● ● ●●● ● ● ● ●●●● ●●●●●●●●● ●●●● ●●●● ● ● ● ●● ● ● ●● ●●●●●●●●●●●●●●●● ●●● ●●●●● ● ●

1.0
● ●●● ● ●●● ●●● ●●●●●● ●●●●●● ●●● ● ●●●●● ●● ●

0.8

0.8
0.8

Probability (Transition)

Probability (Transition)
Probability (Transition)

0.6

0.6
0.6

0.4

0.4
0.4

0.2

0.2
0.2

0.0

0.0
● ● ● ●● ● ●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ●● ●●● ● ● ●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ●● ●
0.0

● ●● ● ● ●●●●●●● ● ●● ●●●● ●● ●● ●● ● ●●● ●●●● ●●●●●● ●● ●● ● ● ●●●

20 40 60 80 100 20 40 60 80 100
20 40 60 80 100
Age Age
Age
Galimany & Getz (Fig. 5) – Forensic Anthropology, Early View
Generating Age Estimates from Trait Data
• TA3 trait data provide significant
age-related information
throughout adulthood
• multiple approaches under
investigation
• TA3 Beta 0.8.5 Software program
(Ousley)
• simplified TA procedure (Getz)
• “traditional” Transition Analysis
framework (Milner & Boldsen)
TA3 Beta
• originally released
at TA3 workshops
(AAFS, AAPA) in
2018

• updated Ver. 0.8.5


released in Oct.
2021 (functionality
is the same as
0.8.0)
Accuracy & Bias of TA3 Beta (Getz & Galimany 2022 - AAFS)
Precision of TA3 Beta (Getz & Galimany 2022 - AAFS)
Sample Size of TA3 Beta
(Getz & Galimany 2022)
• implementation of R package
"randomGLM“ only uses reference
sample skeletons with ALL of the
traits entered from the unknown

• more complete skeletons = greatly


reduced sample size

• age and “population” composition of


the reference sample used to
generate estimate is unknown
Evaluation of TA3 Beta (0.8.5):
Summary of Findings
• Method: TA3 beta does not use “transition analysis” à black box,
machine learning approach
• Accuracy: average accuracy 84.5%
o < 40 years (57.0%, 73/128)
o 40+ years (93.7%, 359/383)
• Precision: interval length correlated to number of traits present
o between 30-40 years with complete or nearly complete skeletons
• Reference Sample: uses only a small subsample of TA3 reference
data to estimate age à composition unknown
Testing TA3 Traits Using a Simplified TA Framework
(Getz & Byrnes 2021 – AAFS Poster) ABH_073

−9
−10
Age Log−likelihood
−11
−12
−13
−14
20 40 60 80 100
Age

clav_med_epiphysisL_01_23 fovea_marginL humerus_lesser_tubercle_lippingR calcaneous_weight

1.0
1.0
● ●● ● ● ●● ●●●●●●●●●●●●●●●● ●●● ●●●●● ● ●
1.0

1.0

● ● ●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●
● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ● ●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●

64

0.8
0.8
0.8

0.8

Probability (Transition)
Probability (Transition)
Probability (Transition)

Probability (Transition)

+ + + +

0.6
0.6
0.6

0.6

additional

0.4
0.4
0.4

0.4

traits

0.2
0.2
0.2

0.2

0.0
●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ●● ●
0.0
0.0

0.0

● ● ● ● ● ●● ● ●●● ●●● ● ●●●● ●●●●●● ● ● ●●● ● ● ● ●● ●● ● ● ● ●●● ●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●● ● ●● ● ●● ● ●● ●

20 40 60 80 100
20 40 60 80 100 20 40 60 80 100 20 40 60 80 100
Age Age Age Age
Accuracy & Precision Using TA3 Traits

uniform prior
distribution artifact
over-estimation

under-estimation

statistical framework for combining multiple traits many traits incorporated into the TA
from pelvis and cranium statistical framework

Sample: Athens (N=201)


Accuracy & Precision (Getz & Byrnes 2021 – AAFS Poster)
TA (Boldsen et al. 2002) TA3 Traits

Accuracy: 73.3% 70.8%


Average Precision: 34.2 yrs. 17.2 yrs.
Athens Sample: Comparison of Commonly Used Methods

Buckberry &
Chamberlain (2002)
Goal!
Brooks & Suchey (1990)

Osborne et al. (2004)


TA
TA3 Traits
(Boldsen et al. 2002)

Lovejoy et al. (1985)


Updated Transition
Analysis Framework
• Milner & Boldsen à updating statistical
framework from Boldsen et al. 2022
with TA3 traits
o Which traits? (more data ≠ better results)
o left or right (both?)
o “population” variation?
o larger sample sizes or sex or population-
specific methods?
o prior distributions?
The Future of TA3?
• larger samples à collaborative
databases, medical imaging
• new statistical and methodological
approaches
• revised trait definitions and
expansion of traits
• better understanding of biological
variation in the aging process and the
expression of skeletal features
Acknowledgements
Collection Managers & Institutions
• Efstratios Valakos and Panagiota Papazafiri—Athens Donated Collection (University of Athens)
• Dawnie Steadman and Lee Jantz—WM Bass Donated Collection (University of Tennessee)
• Erika L’Abbé and Marius Loots—Pretoria Bone Collection (University of Pretoria)
• Pasuk Mahakkanukrauh—Chiang Mai Collection (University of Chiang Mai)
• Susana Garcia—Bocage Museum Collection (Lisbon Natural History Museum)

Funding
• NSF Bio Anth DDRIG #1455810
• NIJ #2014-DN-BX-K007
Dr. Sara M. Getz
Email: smg5036@gmail.com
Website: SaraMGetz.com

You might also like