| Type: | Package |
| Title: | Tests for Same-Source of Toolmarks |
| Version: | 0.0.2 |
| Date: | 2026-09-14 |
| Author: | Jeremy Hadler [aut, cre], Max Morris [ths], Heike Hofmann [ctb] |
| Maintainer: | Jeremy Hadler <hadler13@yahoo.com> |
| Description: | Implements two tests for same-source of toolmarks. The chumbley_non_random() test follows the paper "An Improved Version of a Tool Mark Comparison Algorithm" by Hadler and Morris (2017) <doi:10.1111/1556-4029.13640>. This is an extension of the Chumbley score as previously described in "Validation of Tool Mark Comparisons Obtained Using a Quantitative, Comparative, Statistical Algorithm" by Chumbley et al (2010) <doi:10.1111/j.1556-4029.2010.01424.x>. fixed_width_no_modeling() is based on correlation measures in a diamond shaped area of the toolmark as described in Hadler (2017). |
| Depends: | R (≥ 3.3), plyr (≥ 1.8.4), dplyr (≥ 0.7.2), reshape2 (≥ 1.4.2) |
| Imports: | ggplot2 (≥ 2.2.1) |
| License: | GPL-3 |
| LazyData: | true |
| RoxygenNote: | 6.0.1 |
| NeedsCompilation: | no |
| Packaged: | 2026-09-14 13:09:29 UTC; hadlerjr |
| Repository: | CRAN |
| Date/Publication: | 2026-09-14 21:00:02 UTC |
Toolmark profiles dataset
Description
Data set of toolmarks (profiles) created by screwdrivers under different angles. Tool mark data included here were produced by Prof. Scott Chumbley, Mr. Stephen Davis, Ms. Taylor Grieve, Mr. Ryan Spotts, and Dr. Jeremy Hadler. These data were produced as part of research performed at the Ames Laboratory, located on the Iowa State University campus. Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under Contract No. DE-Ac02-07CH11358.
Usage
ameslab
Format
the dataset consists of a sample of 16 toolmark profiles and descriptors. Toolmarks with the same toolmark identifier are known matches, all other profiles are known non-matches.
- ID
toolmark identifier. Factor variable.
- side
A or B, indicates the side of the screwdriverused to create the toolmark. Factor variable.
- angle
degree under which the toolmark was created.
- rep
replicate number for a toolmark, side, angle composition. Note that the data here provided is not complete. For a more complete data set or more information please contact the references given below.
- profile
list of data sets with one profile each. Measurements are taken at equispaced intervals across the toolmark. .
References
Tool mark data included here were produced by Prof. Scott Chumbley, Mr. Stephen Davis, Ms. Taylor Grieve, Mr. Ryan Spotts, and Dr. Jeremy Hadler. These data were produced as part of research performed at the Ames Laboratory, located on the Iowa State University campus. Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under Contract No. DE-Ac02-07CH11358.
Examples
data(ameslab)
plot(ameslab$profile[[1]]$V1)
chumbley_non_random(ameslab$profile[[1]], ameslab$profile[[2]])
res14 <- fixed_width_no_modeling(ameslab$profile[[1]], ameslab$profile[[4]])
res14$dist_pval
Chumbley Non-Random
Description
This function computes the Chumbley U-Statistic on systemically chosen pairs of windows rather than the original method which selects randomly chosen pairs of windows
Usage
chumbley_non_random(data1, data2, window_opt = 500, window_val = 50,
coarse = 0.25)
Arguments
data1 |
The first tool mark as a 1-column matrix |
data2 |
The second tool mark as a 1-column matrix |
window_opt |
size of the window to be used in the optimization step |
window_val |
Size of the window to be used in the validation step |
coarse |
smoothing parameter for the normalization smooth |
Value
list with
- same_shift_n
Number of same shift offsets used
- diff_shift_n
Number of different shift offsets used
- U
observed U statistic
- p_value
Corresponding p-value
Distance/threshold test for toolmarks
Description
Compute all possible correlations for windows of length n between the class components. Determine the location of the maximized correlation. Given this location, create a diamond around it in the individual matrix of correlations For each offset in this diamond, compute the maximized correlation (1) Determine the distance between the offset for the class and indiviudal components (2) Compute the Threshold test statistics
Usage
fixed_width_no_modeling(dat1, dat2, coarse = 0.25, fine = 0.01,
window.size = 0.6, M = 500)
Arguments
dat1 |
a one column matrix representing a digitized tool mark |
dat2 |
a one column matrix representing a second digitized tool mark |
coarse |
normalization smoothing parameter |
fine |
decomposition smoothing parameter |
window.size |
desired window size for the correlations to compute |
M |
search area restriction |
Value
list with
- max_corr
maximized individual component correlation
- Smooth_offset
optimal Class offset
- Resid_offset
optimal individual offset
- dist_pval
distance p-value
- thresh_pval
threshold p-value
- Above
Number of offsets with correlation bigger than threshold
- total_thresh
2*M+1
- mark1_decomposition
plot of decomposition d1
- mark2_decomposition
plot of decomposition d2
- class_correlations
plot of class correlation
- individual_correlations
plot of individual correlation
- distance_plot
distance plot
- threshold_plot
threshold plot