Working with Misspecified Regression Models

Objectives

Conventional statistical modeling in criminology assumes proper model specification. Very strong and unrebutted criticisms have existed for decades. Some respond that although the criticisms are correct, there is for observational data no alternative. In this paper, we provide an alternative.

Methods

We draw on work in econometrics and statistics from several decades ago, updated with the most recent thinking to provide a way to properly work with misspecified models.

Results

We show how asymptotically, unbiased regression estimates can be obtained along with valid standard errors. Conventional statistical inference can follow.

Conclusions

If one is prepared to work with explicit approximations of a “true” model, defensible analyses can be obtained. The alternative is working with models about which all of the usual criticisms hold.

     

Positive identification of the principal component of a white powder as scopolamine by quantitative one-dimensional and two-dimensional NMR techniques

An unidentified white powder collected as evidence in an intelligence investigation was characterized exclusively by nuclear magnetic resonance (NMR) analysis. A small fraction of the powder dissolved in D2O was subjected to a series of one- and two-dimensional techniques which were used to elucidate the molecular structure of the powder's major component and positively identify it as the scopolamine biotoxin. Quantitative one-dimensional experiments identified individual proton and carbon atom sites, and conventional 14N spectroscopy detected a single nitrogen atom site. Heteronuclear single quantum coherence data correlated all protons to their directly bonded carbon atom, and together with the quantitative spectra, were used to determine the number of protons directly bonded to each carbon atom. The presence of a methyl, carboxyl, and a benzyl group was also identified from these data. Correlation spectroscopy detected a three proton and a nine proton JH,H network, representing a CH2CH moiety and seven carbon atom ring, respectively. These five elements were assembled into an almost complete molecular structure by using long-range, J-coupled, 1H-13C pairs detected by heteronuclear multiple bond correlation (HMBC) spectroscopy and 1H-1H dipolar-coupled pairs found from nuclear Overhauser effect spectroscopy (NOESY) data. Additional oxygen atom sites were inferred from 1H-13C correlation intensities in the HMBC spectra along with 1H and 13C chemical shift values, or directly from NOESY correlations. Only a single oxygen atom site could not be inferred from NMR data, but its presence was inferred from comparisons to target analyte structures to complete the structure of the scopolamine molecule. To confirm these results, an ethanol/H2O solution of the powder was analyzed by direct infusion into an ion trap mass spectrometer. A prominent base signal was observed at m/z 304.1 amu, corresponding to the protonated molecular ion of scopolamine. Subsequently, the ion was selected and subjected to collision-induced dissociation, producing characteristic major MS/MS fragments at m/z 138.1 and 156.1. Comparisons of 1H and 13C chemical shift values and JH,H values measured from our NMR data were found to agree very favorably with previously reported values for scopolamine in D2O.     

The potential for work enforcement: A study of WIN

Given the impediments around them, can welfare recipients be required to work? Most analysts have answered no. They say the recipients are usually kept from employment by socioeconomic barriers, such as insufficient jobs, and the disincentives to work inherent in welfare. Studies of recent AFDC work programs make them look promising but do not directly address the potential for work enforcement. This article, a cross-sectional study of state WIN programs in 1979, suggests that work requirements could raise work levels substantially despite the impediments. But requirements probably do not improve the quality of jobs recipients are able to get. Therefore, enforcement serves the goal of integration, but to achieve greater economic equality will require additional reforms.     

Laws, Models, and Real People: Choice of Chapter in Personal Bankruptcy

It has been a central assumption in bankruptcy policy debates that financially troubled persons faced with bankruptcy will respond to economic incentives and disincentives. Two provisions of the Bankruptcy Code (Chapters 7 and 13) are most commonly used by individual debtors. Under Chapter 7 debtors agree to give up all their property (n excess of state-determined exemptions) to a trustee for sale and distribution to creditors. Under Chapter 13 debtors keep all their property but agree to pay all or part of their debts over three to five years. This empirical study of fifteen hundred consumers in three states explores whether economic incentives and disincentives are in fact the chief factors influencing choice of chapter. The analysis demonstrates that while economic factors play a part, noneconomic factors are also significant, among them intra- and interstate migration, marital status, self-employment, state of residence, and local legal culture. We conclude that to explore fully how individual decisions are made, the simplistic economic model must be replaced by a more sophisticated model that accounts for both economic and noneconomic factors.     

Accuracy of cause of death determination without forensic autopsy examination

Medical examiners and coroners commonly determine cause and manner of death without an autopsy examination. Some death certificates generated in this way may not state the correct cause and manner of death. From the case files of the Department of Forensic Medicine in Sydney, Australia, the authors retrospectively reviewed investigative information of all cases in a 6-month period that were initially considered natural deaths (429). The authors, blinded to autopsy results, accepted 261 cases as appropriate for certification without autopsy and assigned a cause of death to each. Per standard local practice, all cases had been autopsied. The actual causes of death as determined by autopsy were then revealed and compared with the presumed causes of death. Most presumed and actual causes of death were cardiovascular (94% and 80%, respectively). The majority of presumed causes of death were listed as ASCVD as the cases lacked features of a more specific cardiovascular process. A large majority of cases had a presumed cause of death of ischemic heart disease based on individual case details. The actual causes of death demonstrated a large breadth of cardiovascular and noncardiovascular disease processes, even though ischemic heart disease accounted for 62% of deaths. The presumed cause of death was completely wrong in 28% of cases. A nonnatural manner of death was present in 3% of cases. This study demonstrates that experienced forensic pathologists may generate erroneous death certificates for cases that are not autopsied.