Logistic Probability Model Research Articles

The Estimation of Logistic Probability Model of Binomial Data: Simulation Comparison

Properties of various types of estimators of the regression coefficients in linear logistic regression models were considered. The estimators include those based on maximum likelihood, minimum chi-square and weighted least squares. The results of a large scale simulation investigation evaluating the moment properties of the estimators are presented for the case of logistic model with a single explanatory variable.

El Niño–Southern Oscillation and Pacific Decadal Oscillation impacts on precipitation in the southern and central United States: Evaluation of spatial distribution and predictions

Understanding and predicting regional impacts of El Niño–Southern Oscillation (ENSO) and Pacific Decadal Oscillation (PDO) on winter (October–March) precipitation can provide valuable inputs to agricultural and water resources managers. Effects of ENSO and PDO on winter precipitation were assessed in 165 climate divisions throughout the southern United States. A continuous region of significantly (P < 0.05) increased (decreased) winter precipitation in response to El Niño (La Niña) conditions in the preceding summer (June–September Southern Oscillation Index (SOI)) extends across the entire southern United States and as far north as South Dakota. Within this region stronger correlations (r ≤ −0.45) are found along the Gulf of Mexico, southern Arizona, and central Nebraska. Winter precipitation differs significantly (P < 0.1) between warm and cold phase PDO periods only in the south central region, with greatest significance centered in Oklahoma. Enhanced negative La Niña anomalies during PDO cold phases are dominant in the central region (Texas to South Dakota) whereas enhanced positive El Niño anomalies during PDO warm phases are dominant in the southwest (Arizona, Nevada, and California) and southeast (Louisiana to Florida). Validation tests of winter precipitation predictions based on summer SOI and/or PDO‐phase show a decrease of 9% to 16% in the relative Mean Absolute Error (MAE) from the MAE obtained by using the mean as a predictor in areas with strong correlation (r < −0.45) between SOI and precipitation. Logistic regression probability models of having above or below average winter precipitation had up to 77% successful predictions. The advantage of having probabilities of exceeding certain precipitation thresholds at the beginning of a hydrologic year makes logistic regression models attractive for decision makers.

Prospective Budgeting for Home Care

To provide a model for prospective budgeting for home care that is plausible, coherent, flexible, and sufficiently tractable that it can serve as a template for practical decision making and to clarify what would be the data requirements and statistical framework to calibrate the model. Methods used are standard risk-neutral expected value theory, cost benefit analysis, and the conditional logistic probability model. A simple but effective prospective budgeting model that provides analytic scaffolding for a practical decision support system for home care case managers, consultants, and program evaluators that can improve program equity, efficiency, and effectiveness. The author criticizes the well-known Titration Budgeting Model of Weissert, Chernew, and Hirth in terms of its logical and operational problems but then goes on to develop a framework within which the goals of the titration model can be met and home care resources can be more efficiently allocated.

Intoxicated motor vehicle passengers warrant screening and treatment similar to intoxicated drivers.

Alcohol interventions decrease alcohol consumption and recurrent injury. The study hypotheses are (1) intoxicated passengers are similar to intoxicated drivers in crashes and driving under the influence of alcohol (DUI), and (2) DUI conviction rates after injury are low. Intoxicated motor vehicle occupants hospitalized for injury in 1996-1998 were matched to the state traffic database for crashes and DUI. Drivers and passengers were compared for crashes and DUI in the 2 years preceding and 1 year after admission. Driver DUI citation at the time of admission was also recorded. A logistic regression model for crash and DUI probability was constructed. Six hundred seventy-four patients met inclusion criteria. In the 2 years preceding admission, passengers and drivers were equally cited for crashes (14.7% vs 19.3%, p = 0.12). In 1 year after admission, they were also equally cited (7.1% vs 7.7%, p = 0.92). Driver/passenger status was not a predictor by logistic regression; 13.4% of intoxicated drivers were convicted of DUI for the admitting crash. Intoxicated passengers and drivers are equally likely to be cited for crashes and DUI before and after admission for injury. Few admitted intoxicated drivers are convicted of DUI. Screening and intervention for all intoxicated crash occupants is warranted.

Labor Migration and Earnings Differences: The Case of Rural China

This article presents empirical evidence supporting the hypothesis that the decision of Chinese rural people to migrate is affected by noneconomic forces. Using the 1820 rural household survey from Sichuan province migration behavior and earnings are analyzed. Logistic probability models show that the decision to migrate is driven largely by the availability of rural non-farm employment opportunities. However if jobs are available in the province even at lower wages rural people will choose to stay rather than to migrate. The major deterrent to migration is the lack of safety during transportation and in destination cities as well as forced separation from families. Moreover rural migrants are denied the legal right to reside permanently in cities. Furthermore findings reveal that although labor productivity in migrant activities is higher than it is in local non-farm sectors the economic cost of migration in China is so high as to significantly limit reallocation. One implication of this study is that without artificial barriers to labor migration the number of migratory workers would be higher. Since the important objective of the Chinese government is to control rural-to-urban migration it is likely that restrictive migration policies will continue. However putting aside the issue of social justice the economic efficiency loss under the Chinese system should not be ignored.

Colonial breeding imposes increased predation: experimental studies with herons

Experiments were carried out over 2 consecutive years to test whether heron nest predation differs between colonial and non-colonial nests. Dummy nests containing chicken eggs dyed light blue were placed in a mixed Night Heron and Little Egret heronry and in a control copse without nesting herons; in both areas Hooded Crows were the main egg predators. Experiments were performed both when the true nests had eggs (synchronized nests) and when the true nests contained chicks (delayed nests). Dummy nests within the heronry were more preyed upon than isolated dummy nests. Delayed dummy nests suffered higher predation than synchronized ones. A predictive logistic regression model for predation probability was developed for the dummy nests placed in the centre of the colony; for the synchronized dummy nests the distance from the nearest occupied real nest entered the model and dummy nests close to real nests were less likely to be preyed upon. Nesting in a colony is costly in terms of predation probability, pro...

Evaluation of two biological markers of tobacco exposure

This study assesses serum thiocyanate and expired-air carbon monoxide based on findings for 8,895 participants in the Multiple Risk Factor Intervention Trial. For each marker and combination of markers, a logistic probability model was used to generate a plot of true positive vs false positive rates (an operating characteristic curve), an optimal cutpoint, and the positive predictive value for discriminating tobacco smokers (cigarettes, pipes, cigars, or cigarillos) from non-tobacco smokers. Optimal cutpoints were 83 μmol/liter for thiocyanate and 10 ppm for carbon monoxide. These cutpoints were applied to the entire group to evaluate accuracy. Specificity was higher for carbon monoxide than for thiocyanate (91.8 vs 87.3%) and sensitivity was lower (85.7% for carbon monoxide vs 88.8% for thiocyanate). When carbon monoxide and thiocyanate were both used, specificity was 92.7% and sensitivity was 89.1%. The estimated cutpoints derived from the Multiple Risk Factor Intervention Trial result in predictive values greater than 87% when the prevalence of smokers is greater than 50%. Cutpoints should be modified to maximize the predictive value when the prevalence of smokers is less than 50%. Both serum thiocyanate and carbon monoxide are good markers of smoking exposure. Serum thiocyanate may be preferable to carbon monoxide because of its longer half-life.

An analysis of effects of fetal order and parental age on retinoblastoma

78 cases of retinoblastoma (Rb) diagnosed at the Eye ENT Hospital, Shanghai from 1953 to 1985 were studied. The results of multiple Logistic regression analysis and conditional probability model showed that high fetal order had a moderate risk (OR = 1.28, P less than 0.05) of developing unilateral Rb. Both unilateral and bilateral cases were independent of parental ages, and no interaction could be found between fetal order and paternal or maternal age; also, the effect of fetal order was not confounded by paternal ages.

Race, moving status, and urban services in central cities

The influences of race and moving status on the likelihood of receiving unsatisfactory or inadequate urban services are investigated using logistic probability models. The analysis bears upon the perennial issue of felt injustice and the charge that white America has a policy of sustaining separate and unequal societies for minorities in cities. The results also speak to the limits on market adjustments to correct undesirable public outcomes. Statistically significant racial distinctions are isolated in the assessment of five urban services; in three cases these are unambiguously consequential. The extent to which moving costs and limited housing opportunities inhibit housing market adjustments to uneven service provision is measured by a decomposition of estimated probabilities for minorities and whites. The evidence generated supports continuing policies that expand minority housing opportunities and reduce impediments to the mobility of minorities in urban areas.

Efficiency considerations in the analysis of a competing‐risk problem

AbstractIn many survival studies involving more than one failure mode the data collected include the failure times as well as the failure types. An analysis of these data for evidence of a treatment effect based on a logistic probability model for failure type sacrifices precision for simplicity. ignoring the information contained in the failure times. Alternatively, a competing‐risk analysis using a cause‐specific hazard model may make more efficient use of the available information, but involves additional assumptions and the possibility that the model selected may not be entirely correct. Aspects of this tradeoff between efficiency and simplicity are considered in the context of a two‐sample competing‐risk problem involving proportional hazards for the incidence of failure attributable to each of two different causes.

Worker mobility and reservation wages for minority youth job seekers

Job offers requiring varying commuting times to work from inner-city areas and the reservation wages for a sample of young Chicano job seekers are examined. A logistic probability model of job acceptance using the minimum wage as an acceptance-rejection cutoff in relation to the reservation wage data is estimated by maximum likelihood techniques and suggests that the level of schooling is an important determinant of the probability of job acceptance. The results of this analysis suggest a restriction on the geographic range of job acceptances to the extent that minimum wage job offers located away from inner city, minority neighborhoods would most likely be rejected by the sample of minority youth job seekers.

Understanding and predicting the yield in the MBA admissions process

This paper examines the factors influencing theyield — the proportion of accepted applicants that confirm their intention to attend — in the admissions process for the MBA program of the school of management at a large metropolitan university. A stochastic model of a student's MBA program choice decision is presented. Using the resulting logistic probability model, the authors examine the impact of student traits (both aptitude-related and non-aptitude-related) and program characteristics, known to admissions officials, on the probability of confirmation. The results of the analysis provide a basis for measuring the dimensions of the market for educational services this MBA program provides, suggest a strategic response to this market, and offer a way to measure the profitability of various strategic response options. In particular, the results demonstrate that financial aid awards can increase dramatically the probability of confirmation and that confirmation probabilities are affected significantly by non-aptitude-related student characteristics.

Specification error in multinomial logit models

In this article, we analyze the omitted variable bias problem in the multinomial logistic probability model. Sufficient, as well as necessary, conditions under which the omitted variable will not create asymptotically biased coefficient estimates for the included variables are derived. Conditional on the response variable, if the omitted explanatory and the included explanatory variable are independent, the bias will not occur. Bias will occur if the omitted relevant variable is independent with the included explanatory variable. The coefficient of the included variable plays an important role in the direction of the bias.

The Risk of Personal Victimization Among the Aged

This study examines the likelihood of personal victimization among the aged as a function of a multitude of factors such as characteristics of the victim and environmental conditions. Although the effects of a number of variables (e.g., age, sex, race, and income) on victimization have been studied, they are generally restricted to descriptive or bivariate analyses. Attempts to examine the impact of all relevant factors simultaneously have been rare. This is probably due to the small proportion of crime victims included in most victimization surveys and the fact that statistical techniques dealing with discrete dependent variables are not yet widely used by researchers. In this study, a maximum likelihood logistic probability model was applied to delineate the effects of individual and environmental attributes on victimization. Data for this analysis came from the National Crime Survey conducted between 1973 and 1976 by the Census Bureau for the Law Enforcement Assistance Administration. Logit analysis revealed that characteristics of both the victim and the environment have fairly consistent effects on the likelihood of personal victimization.

The levels and determinants of household insulation activity

This paper provides evidence of current levels of household insulation, including storm windows and storm doors. A logistic probability model is used to determine the effect of heating fuel price, climate, and household characteristics (such as age of head and income) on the likelihood that a household will be well insulated or will intend to add insulation in the near future. Households headed by individuals under 50 years of age are better insulated, as are households with annual income greater than $20,000. Middle income families ($15,000–30,000/yr) are the most likely to plan to add insulation, while low income families are least likely. Households with a head over 65 years of age are less likely to plan to add insulation. These findings, in addition to evidence concerning households which are currently poorly insulated and do not plan to add insulation, are used to recommend modification of recently passed energy legislation.

Enrollment choices in different types of HMOs: a multivariate analysis.

Enrollment decisions of a sample of an employed population choosing among open-panel and closed-panel HMOs and Blue Cross/Blue Shield are analyzed. This report, unlike previous ones, overcomes some of the difficulties of bivariate analysis by the use of the multivariate logistic probability model, logit. The results show that there are four consistent predictors of enrollment choice: previous source of care as the measure of access; family life stage and chronic conditions per family member as indicators of health risk; per capita income as the measure of economic vulnerability; and health concern. Having a private physician as the source of care is the best single predictor, its absence predicting a higher probability of enrollment in the closed, and its presence in the open-panel HMO. Higher risk life stage families, younger and with more children, are more likely to join the open-panel plan than the closed or retain BC/BS; higher incomes and larger numbers of chronic conditions appear to have the same effects. Higher levels of health concern, on the other hand, predict a greater probability of choosing the closed-panel plan. The probability of enrollment in any HMO is predicted with more than 50 per cent accuracy for 60 per cent of the sample. Choice between open and closed-panel plans is predicted with an accuracy in excess of 50 per cent for 80 per cent, and with an accuracy greater than 90 per cent for over 10 per cent of potential enrollees. The applicability of this approach to HMO feasibility analysis and planning is clearly indicated.