2. Matchsticks
Three functional classes
i. Safety – purpose built striking surface
ii. Strike anywhere – no purpose built striking surface
iii. Waterproof – water resistant
Many different commercial brand manufacturers
2
3. Matchstick Forensics
Stereomicroscopy SEM-EDS X-ray Diffraction
Visually
different?
Y/N
Chemically
different?
Y/N
Structurally
different?
Y/N
Qualitative similarity
assessment on Q vs. K
3
4. Matchstick Forensics
Stereomicroscopy SEM-EDS X-ray Diffraction
Visually
different?
Y/N
Chemically
different?
Y/N
Structurally
different?
Y/N
Qualitative similarity
assessment on Q vs. K
How can we improve
trace evidence analysis?
4
24. ANOSIM
PCA emphasises differences between sample groups
Are the emphasised differences statistically significant?
𝑅 =
𝑟𝐵 − 𝑟 𝑊
𝑛(𝑛 − 1)/4
Test Statistic R
−1 ≤ R ≤ 1
R→1 Differences between
groups become more significant
24
27. Data Pretreatment
Matchstick Multivariate Data Set Incorporating Physical and Chemical
Data
Brand
B
11
Mg
25
Fe
57
Zn
(all)
Zn
70
Binder
P/A
NC
Binder
Alkyd
Binder
Shellac
Binder
Head
Structure
Paper
Splint
REI 0.1 4.0 0.0 1.4 0.2 1 0 1 0 2 0
REI 0.1 3.9 0.0 1.3 0.1 1 0 1 0 2 0
REI 0.1 4.6 0.0 1.3 0.2 1 0 1 0 2 0
REI 0.0 4.2 0.0 1.2 0.1 1 0 1 0 2 0
Matchstick Multivariate Data Set Incorporating Physical and Chemical
Data
All Variable Values Standardized by Maximum Value in each Variable
Brand
B
11
Mg
25
Fe
57
Zn
(all)
Zn
70
Binder
P/A
NC
Binder
Alkyd
Binder
Shellac
Binder
Head
Structure
Paper
Splint
REI 0.1 2.9 0.0 0.5 0.5 100 0 100 0 100 0
REI 0.1 2.8 0.0 0.5 0.5 100 0 100 0 100 0
REI 0.1 3.3 0.0 0.5 0.5 100 0 100 0 100 0
REI 0.1 3.1 0.0 0.4 0.5 100 0 100 0 100 0
Matchstick Multivariate Data Set Incorporating Physical and Chemical
Data
All Variable Values Transformed by LN(X+1)
Brand
B
11
Mg
25
Fe
57
Zn
(all)
Zn
70
Binder
P/A
NC
Binder
Alkyd
Binder
Shellac
Binder
Head
Structure
Paper
Splint
REI 0.1 1.3 0.0 0.4 0.4 4.6 0.0 4.6 0.0 4.6 0
REI 0.1 1.3 0.0 0.4 0.4 4.6 0.0 4.6 0.0 4.6 0
REI 0.1 1.5 0.0 0.4 0.4 4.6 0.0 4.6 0.0 4.6 0
REI 0.1 1.4 0.0 0.4 0.4 4.6 0.0 4.6 0.0 4.6 0
27
28. Gower dissimilarity measure
Dissimilarity Analysis
|𝑦1𝑗 − 𝑦2𝑗 |
(𝑚𝑎𝑥𝑗 − 𝑚𝑖𝑛𝑗 )
𝑝
𝑗=1
Assesses both discrete and
continuous variable
contributions to similarity
between samples
𝑤12𝑗 𝑠12𝑗
𝑝
𝑗=1
𝑤12𝑗
𝑝
𝑗=1
28
29. NMDS
Displays sample relationships in multivariate space
Based on calculated dissimilarities
MDS for metric dissimilarity measure
NMDS for non-metric dissimilarity measure
A B
C
D(A, B) ≥ 0
D(A, B) = 0 ⟷ A=B
D(A, B) = D(B, A)
D(B, C) ≥ [D(A, B) + D(C, A)]
29
31. Forensic Signature
ANOSIM R values
W vs. S 0.609
W vs. SA 0.981
S vs. SA 0.431
ANOSIM R values
R=1 for all
Commercial Brand
pairwise comparisons
31
33. Summary
Combination of analyses excellent for classification and
brand determination
Multivariate statistical analysis confirms hypothesis
Very visual and intuitive final result
33
34. Discussion
How do we improve Trace Evidence Analysis?
Matchsticks are a good model
Branching out: fibers, paint, glass
Sufficient sample sizes
Automated statistical methods
34
35. Acknowledgements
Mr. Chad Schennum – Forensic Scientist1
Mr. Tomson Huynh – FTIR Assistant2
Mr. Spencer Nwogoku – Laboratory Assistant2
Dr. Joseph Turner – Director of Instrumentation2
Mr. Thomas Pugh – Introducing Presenter2
1VA Department of Forensic Science
2Virginia Commonwealth University
35