Likelihood Ratios for Deep Neural Networks in Face Comparison

In this study, we aim to compare the performance of systems and forensic facial comparison experts in terms of likelihood ratio computation to assess the potential of the machine to support the human expert in the courtroom. In forensics, transparency in the methods is essential. Consequently, state-of-the-art free software was preferred over commercial software. Three different open-source automated systems chosen for their availability and clarity were as follows: OpenFace, SeetaFace, and FaceNet; all three based on convolutional neural networks that return a distance (OpenFace, FaceNet) or similarity (SeetaFace). The returned distance or similarity is converted to a likelihood ratio using three different distribution fits: parametric fit Weibull distribution, nonparametric fit kernel density estimation, and isotonic regression with pool adjacent violators algorithm. The results show that with low-quality frontal images, automated systems have better performance to detect nonmatches than investigators: 100% of precision and specificity in confusion matrix against 89% and 86% obtained by investigators, but with good quality images forensic experts have better results. The rank correlation between investigators and software is around 80%. We conclude that the software can assist in reporting officers as it can do faster and more reliable comparisons with full-frontal images, which can help the forensic expert in casework.     

Homelessness Interventions in Georgia: Rapid Re-Housing,Transitional Housing,and the Likelihood of Returning to Shelter

Since 1987, billions of dollars in homeless assistance have been allocated annually by the U.S. federal government. Yet few evaluations of homelessness interventions exist. This study analyzes the likelihood that households in Georgia returned to shelter within two years of leaving one of three interventions: rapid re-housing (RRH), transitional housing (TH), and emergency shelter (ES), with the latter serving as a reference. Using propensity scores, RRH households were matched to comparable TH and ES households. Generalized linear mixed modeling then controlled for household characteristics as well as variation between intervention implementations. We find that the likelihood of returning to shelter did not seem to be affected by whether study households were gradually transitioned or rapidly placed into housing. Additionally, the effect of TH for households without children seems highly dependent on the intervention’s implementation, which deserves further study. Our findings are generalizable to a small, better resourced segment of the general homeless population.     

Governmental Accounting Maturity Toward IPSASs and the Approximation to National Accounts in the European Union

International Public Sector Accounting Standards (IPSASs) are a good reference for a harmonized microaccounting system allowing more transparency and quality in public sector accounting across EU member-States. However, questions remain concerning IPSASs contribution to the convergence between Governmental Accounting (GA) and the National Accounts (NA). This article assess how the proximity to an IPSAS-based accounting system in GA has impact on the diversity and materiality of GA-NA budgetary deficit/surplus adjustments, hence analyzing whether IPSASs might contribute to GA-NA reconciliation. Main findings show that IPSASs do not make considerable difference in terms of GA-NA adjustments, so IPSASs-based EPSASs will hardly contribute to approaching GA-NA.     

Validation of the 4‐aminophenol color test for the differentiation of marijuana‐type and hemp‐type cannabis

The analysis of cannabis plant material submitted to seized‐drug laboratories was significantly affected by the signing of the Agricultural Improvement Act of 2018, which defined hemp and removed it from the definition of marijuana in the Controlled Substances Act. As a result, field law enforcement personnel and forensic laboratories now are in need of implementing new protocols that can distinguish between marijuana‐type and hemp‐type cannabis. Colorimetric tests provide a cost‐effective and efficient manner to presumptively identify materials prior to submission to a laboratory for analysis. This work presents the validation of the 4‐aminophenol (4‐AP) color test and demonstrates its utility for discriminating between marijuana‐type and hemp‐type cannabis (i.e., typification). Validation studies included the testing of numerous cannabinoid reference materials, household herbs, previously characterized cannabis plant samples, and real‐case samples. The 4‐AP test reliably produces a pink result when the level of Δ⁹‐tetrahydrocannabinol (THC) is approximately three times lower than the level of cannabidiol (CBD). A blue result is generated when the level of THC is approximately three times higher than that of CBD. Inconclusive results are observed when the levels of THC and CBD are within a factor of three from each other, demonstrating the limitations of the test under those scenarios.     

Allele frequencies data and statistic parameters for 13 STR loci in a population of the Brazilian Amazon Region

Allele frequencies for 13 short tandem repeat (D3S1358, vWA, D21S11, D18S51, D5S818, D13S317, D7S820, TH01, TPOX, D16S539, CSF1PO, D8S1179 and FGA) loci were determined in a sample of 325 unrelated individuals from the population of the Amazon of Belém, Brazil. These loci are the most commonly used in forensic and paternity testing. The forensic parameters investigated presented high values. The power of discrimination and the probability of exclusion for these 13 STRs are 99.999999999992% and 99.9998%, respectively. In conclusion, these 13 markers are suitable for forensic analysis and paternity tests of the Amazonian population.     

Airway obstruction by laryngeal masses in sudden and homicidal deaths

We present two cases of autopsy detection of laryngeal masses with medico-legal implications. The first is a 56-year-old man who died suddenly of asphyxia due to upper airway obstruction caused by a large glosso-epiglottic retention cyst and glottic oedema. Hypothesis of medical liability was raised, due to delayed tracheotomy as the result of repeated failed attempts at oral intubation by various physicians. Difficult oral intubation due to the presence of a laryngeal mass in an asphyxiating subject requires rapid tracheotomy. The second case deals with a 43-year-old woman who died of asphyxia due to airway obstruction, as a result of injury of a cavernous laryngeal haemangioma after homicidal manual strangulation, with severe haemorrhagic infiltration of the surrounding soft tissues. Damage to laryngeal/hypopharyngeal masses should be considered as a possible sign of manual strangulation, as well as neck skin bruises/abrasions and laryngeal haemorrhages.     

Oil and the duration of dictatorships

Theoretical models do not reach an unambiguous conclusion concerning the effects of natural resource endowment on the duration of dictatorial regimes. We assess empirically, for the first time, the relationship between oil endowment and the duration of autocratic leaders. Using a dataset comprising information for 106 dictators, our empirical analysis indicates that dictators in countries which are relatively better endowed in terms of oil tend to stay longer in office. The result is robust to changes in the definition of dictatorial regimes and in the specifications used in the econometric analysis.