The area under the receiver operating characteristic (ROC) curve, known as the AUC, is currently considered to be the standard method to assess the accuracy of predictive distribution models. It avoids the supposed subjectivity in the threshold selection process, when continuous probability derived scores are converted to a binary presence‐absence variable, by summarizing overall model performance over all possible thresholds. In this manuscript we review some of the features of this measure and bring into question its reliability as a comparative measure of accuracy between model results. We do not recommend using AUC for five reasons: (1) it ignores the predicted probability values and the goodness-of-fit of the model; (2) it summarises the test performance over regions of the ROC space in which one would rarely operate; (3) it weights omission and commission errors equally; (4) it does not give information about the spatial distribution of model errors; and, most importantly, (5) the total extent to which models are carried out highly influences the rate of well-predicted absences and the AUC scores.

/pdf/auc-a-misleading-measure-of-the-performance-of-predictive-qqj6tbb550.pdf

AUC: a misleading measure of the performance of predictive distribution models

\s=b\To ascertain the prevalence of posttraumatic stress disorder (PTSD) and risk factors associated with it, we studied a random sample of 1007 young adults from a large health maintenance organization in the Detroit, Mich, area. The lifetime prevalence of exposure to traumatic events was 39.1%. The rate of PTSD in those who were exposed was 23.6%, yielding a lifetime prevalence in the sample of 9.2%. Persons with PTSD were at increased risk for other psychiatric disorders; PTSD had stronger associations with anxiety and affective disorders than with substance abuse or dependence. Risk factors for exposure to traumatic events included low education, male sex, early conduct problems, extraversion, and family history of psychiatric disorder or substance problems. Risk factors for PTSD following exposure included early separation from parents, neuroticism, preexisting anxiety or depression, and family history of anxiety. Life-style differences associated with differential exposure to situations that have a high risk for traumatic events and personal predispositions to the PTSD effects of traumatic events might be responsible for a substantial part of PTSD in this population. (Arch Gen Psychiatry. 1991 ;48:216-222)

Traumatic events and posttraumatic stress disorder in an urban population of young adults

Initial human and financial capital as predictors of new venture performance

ContextAlthough diabetes mellitus is one of the most prevalent and costly chronic
diseases in the United States, no estimates have been published of individuals'
average lifetime risk of developing diabetes.ObjectiveTo estimate age-, sex-, and race/ethnicity-specific lifetime risk of
diabetes in the cohort born in 2000 in the United States.Design, Setting, and ParticipantsData from the National Health Interview Survey (1984-2000) were used
to estimate age-, sex-, and race/ethnicity-specific prevalence and incidence
in 2000. US Census Bureau data and data from a previous study of diabetes
as a cause of death were used to estimate age-, sex-, and race/ethnicity-specific
mortality rates for diabetic and nondiabetic populations.Main Outcome MeasuresResidual (remaining) lifetime risk of diabetes (from birth to 80 years
in 1-year intervals), duration with diabetes, and life-years and quality-adjusted
life-years lost from diabetes.ResultsThe estimated lifetime risk of developing diabetes for individuals born
in 2000 is 32.8% for males and 38.5% for females. Females have higher residual
lifetime risks at all ages. The highest estimated lifetime risk for diabetes
is among Hispanics (males, 45.4% and females, 52.5%). Individuals diagnosed
as having diabetes have large reductions in life expectancy. For example,
we estimate that if an individual is diagnosed at age 40 years, men will lose
11.6 life-years and 18.6 quality-adjusted life-years and women will lose 14.3
life-years and 22.0 quality-adjusted life-years.ConclusionsFor individuals born in the United States in 2000, the lifetime probability
of being diagnosed with diabetes mellitus is substantial. Primary prevention
of diabetes and its complications are important public health priorities.

/pdf/lifetime-risk-for-diabetes-mellitus-in-the-united-states-3ldpmi6ejj.pdf

Lifetime Risk for Diabetes Mellitus in the United States

Objectives. This 11-city study sought to identify risk factors for femicide in abusive relationships. Methods. Proxies of 220 intimate partner femicide victims identified from police or medical examiner records were interviewed, along with 343 abused control women. Results. Preincident risk factors associated in multivariate analyses with increased risk of intimate partner femicide included perpetrator’s access to a gun and previous threat with a weapon, perpetrator’s stepchild in the home, and estrangement, especially from a controlling partner. Never living together and prior domestic violence arrest were associated with lowered risks. Significant incident factors included the victim having left for another partner and the perpetrator’s use of a gun. Other significant bivariate-level risks included stalking, forced sex, and abuse during pregnancy. Conclusions. There are identifiable risk factors for intimate partner femicides.

/pdf/risk-factors-for-femicide-in-abusive-relationships-results-2qrfqxfojj.pdf

Risk Factors for Femicide in Abusive Relationships: Results From a Multisite Case Control Study

Several test statistics are proposed for the purpose of assessing the goodness of fit of the multiple logistic regression model. The test statistics are obtained by applying a chi-square test for a contingency table in which the expected frequencies are determined using two different grouping strategies and two different sets of distributional assumptions. The null distributions of these statistics are examined by applying the theory for chi-square tests of Moore Spruill (1975) and through computer simulations. All statistics are shown to have a chi-square distribution or a distribution which can be well approximated by a chi-square. The degrees of freedom are shown to depend on the particular statistic and the distributional assumptions. The power of each of the proposed statistics is examined for the normal, linear, and exponential alternative models using computer simulations.

Stanley Lemesbow

Papers

Goodness of fit tests for the multiple logistic regression model