Table 1.
Summary of STRmix v2.6 and EFM v2.1.0 interpretation parameters and reported LR values.
Fig 1.
Schematic overview of the Likelihood Ratio (LR) system, adapted from [47, 48].
Table 2.
Summary of the total number of PROVEDIt mixture profiles and H1-true and H2-true propositions analyzed in both STRmix and EFM for 2P, 3P, and 4P mixtures.
Fig 2.
Distribution of log10 (LR) values for H1-true and H2-true tests assessed by STRmix and EFM for two, three, and four person mixtures.
The x-axis shows the labels of propositions (H1 and H2), software (STRmix and EFM), and the NOC = 2 Person, 3 Person, and 4 Person. LRs are plotted on the y-axis as log10(LR) values. All samples from different mixture ratios, total DNA template amounts, and DNA treatments are built into this global/overall distribution plot. The plot contains a total of 308 H1-true tests and 308 H2-true tests for the 2P analysis, 441 H1-true and 441 H2-true calculations for the 3P analysis, and 508 H1-true and 508 H2-true tests for the 4P mixtures. STRmix provides an LR value of 0 for excluded loci resulting in profile LR of 0, while EFM gives a non-zero LR value (generally very close to zero). Profiles with LR results of 0 from STRmix are plotted at −125 on the log10 scale. ** Two H1-true test interpretations of 2P mixtures for which STRmix assigned profile LRs of 0 (plotted at H1 true STRmix NOC = 2 Person in magenta at −125 on the log10 scale and discussed in detail in Section 3.6).
Fig 3.
Summary of adventitious exclusionary and inclusionary support from both LR systems with their corresponding log10(LR) values.
The x-axis shows the log10(LR) values for these adventitious exclusionary and inclusionary cases. The y-axis shows the labels of the tested propositions (H1 and H2) from each software (STRmix and EFM). ** in (A.) are the two 2P H1-true test interpretations for which STRmix assigned profile LRs of 0 (plotted in magenta at–Infinity (-Inf) on the log10 scale and discussed in detail in Section 3.6).
Table 3.
Summary of the number of observations and frequency (%) of known contributor analyses (H1-true tests) and known non-contributor analyses (H2-true tests) that yielded log10(LR) values > 0 (or LR > 1) and log10(LR) values < 0 (or LR <1), respectively.
Fig 4.
Empirical ROC plots used to study discrimination performance of the LR systems.
ROC plots are built per varying NOC and software. Each NOC dataset is composed of profiles of different DNA quality, quantity, and mixture proportions. The red, blue, and green curves are the ROC plots constructed using LR values of known contributors and known non-contributors of 2P, 3P, and 4P mixtures analyzed within STRmix, respectively. ROC plots constructed with LR values assigned by EFM are shown in orange (2P), cyan (3P), and magenta (4P). The plot contains a total of 308 H1-true tests and 308 H2-true tests for the 2P analysis, 441 H1-true and 441 H2-true calculations for the 3P analysis, and 508 H1-true and 508 H2-true tests for the 4P mixtures. The calculated AUCs and p-values are shown. All p-values were > 0.05.
Fig 5.
Global overall profile H1-true test and H2-true test log10(LR) values assigned by STRmix and EFM.
** in (A.) are the two 2P H1-true test interpretations for which STRmix assigned profile LRs of 0 (plotted in magenta at–Infinity (-Inf) on the log10 scale and discussed in detail in Section 3.6).
Fig 6.
Relative frequency histograms of the degree of differences in log10(LR) values between the two LR systems.
The absolute difference in log10(LR) are shown on the y-axis. The square bracket “[" in the interval notation "[)" indicates that the endpoint is included in the interval and the parenthesis ")" in the interval notation "[)" indicates that the endpoint is not included. For example, [1, 2], is the interval of values between 1 and 2, including 1 and up to but not including 2, i.e., 1 ≤ values < 2. The x-axis shows the relative frequencies (in %) of the differences in log10(LR) values between the LR systems occurring within each bin. The relative frequencies are also labeled above each bar of the histogram. ** are the two 2P H1-true test interpretations for which STRmix assigned profile LRs of 0 (binned into the [17, Inf) category and discussed in detail in Section 3.6).
Fig 7.
Distribution of differences in log10(LRs) across major, minor, and equal contributors.
The differences in log10(LRs) here shown between EFM and STRmix (log10(LR)EFM−log10(LR)STRmix) are plotted on the x-axis in log10 scale. The y-axis shows the labels of the types of POI with their corresponding number of observations. ** are the two 2P H1-true test interpretations for which STRmix assigned profile LRs of 0 (plotted in magenta at Infinity (Inf) on the log10 scale and discussed in detail in Section 3.6).
Fig 8.
Concordance/discordance tables of the binned LR values assigned by STRmix and EFM into their verbal equivalents.
The tables display the results of the categorization of the LRs for both the H1-true tests of (A) 2P where ** are the two 2P H1-true test interpretations for which STRmix assigned profile LRs of 0 (binned into the Exclusionary category and discussed in detail in Section 3.6), (B) 3P, and (C) 4P and H2-true tests of (D) 2P, (E) 3P, and (F) 4P generated in STRmix and EFM into their corresponding verbal expression. Also, the tables demonstrate the observed differences in the verbal expressions between the two LR systems. The number of cases that resulted in same verbal expression between STRmix and EFM fell inside the diagonal (white cells). All the numbers outside the diagonal (shaded cells) are indication of cases where LRs from both software were classified into different categories and resulted in shifting by one or more than one verbal category (indicated by different shades as shown by the legend). Values in and above the diagonal are the results of the verbal expression of LRs produced in EFM while values in and below the diagonals are the results of the verbal expression of LR values assigned by STRmix. The verbal expressions are shown at the top and left edges of the tables.