Proportional-odds Logit Models Research Articles

Markov switching stereotype logit models for longitudinal ordinal data affected by unobserved heterogeneity in responding behavior

AbstractWhen asked to assess their opinion about attitudes or perceptions on Likert-scale, respondents often endorse the midpoint or extremes of the scale and agree or disagree regardless of the content. These responding behaviors are known in the psychometric literature as middle, extremes, aquiescence and disacquiescence response styles that generally introduce bias in the results. One of the key motivations behind our approach is to account for these attitudes and how they evolve over time. The novelty of our proposal, in the context of longitudinal ordered categorical data, is in considering simultaneously the temporal dynamics of the responses (observable ordinal variables) and unobservable answering behaviors, possibly influenced by response styles, through a Markov switching logit model with two latent components. One component accommodates serial dependence and respondent’s unobserved heterogeneity, the other component determines the responding attitude (due to response styles or not). The dependence of the responses on covariates is modelled by a stereotype logit model with parameters varying according to the two latent components. The stereotype logit model is adopted because it is a flexible extension of the proportional odds logit model that retains the advantage of using a single parameter to describe a regressor effect. In the paper, a new interpretation of the parameters of the stereotype model is given by defining the allocation sets as intervals of values of the linear predictor that identify the most probable response. Unobserved heterogeneity, serial dependence and tendency to response style are modelled through our approach on longitudinal data, collected by the Bank of Italy.

Modelling urban freight generation using linear regression and proportional odds logit models

Receivers and intermediary establishments typical to urban areas have unique freight behaviour and are understudied, particularly in developing countries. This paper uses establishment-based freight survey data to examine their freight generation patterns in the Chennai metropolitan area, India. Three sets of Simple Linear Regression (SLR) models, three single predictor proportional odds logit models, one multiple linear regression model, and one multiple predictor proportional odds logit model with business size (employment, area, and operational age) and indicators of establishment category as regressors are developed each for both freight production and freight attraction. Partial proportional odds logit models are developed in a few cases to overcome the limitations of proportional odds logit models. The best SLR model varied with the establishment category for freight production and attraction. The establishment area model is the best among single predictor proportional odds logit models. The multiple predictor proportional odds logit models marginally improved the fit over single predictor models. The proportional odds logit model results show that the establishment category has a greater impact on freight generation levels than business size variables. Since earlier studies rarely focused on receivers and intermediary establishments, policymakers may benefit from the developed models and study insights while estimating freight demand and developing freight policies.

주택연금 가입자의 노후 대책 충분성에 대한 연구

This study uses a proportional odds logit model to find out the factors that affect the 'sufficiency of old-age measures prepared so far' for reverse mortgage users among the Housing Finance Corporation's “2021 Housing Pension Demand Survey”. Utilized. Through this, we tried to find an alternative to stabilizing the old-age income of the elderly, targeting the reverse mortgage subscribers. As a result of the analysis, the most influential factor on the sufficiency of reverse mortgage subscribers' retirement plans was the sufficiency of average monthly income, secondly the sufficiency of assets for living in their old age, thirdly satisfaction with reverse mortgage. And next, area of residence followed by average monthly income. The demographic characteristics of the reverse mortgage subscribers had no significant effect at all on the 'sufficiency of old-age measures prepared so far' except for the region. The reverse mortgage subscribers commonly receive monthly payments through a mortgage, so it can be seen that the average monthly income is more important than anything else. It was clearly derived from the analysis results of this study that the monthly income or pension paid each month is the most suitable means of economic stability in old age to prepare for the longevity risk of the reverse mortgage subscribers.

정책모기지 이용자 만족도 영향 요인 연구

Policy mortgage is a policy financial product that enables long-term fixed interest rate, low interest rate, long-term loan period, and installment repayment for the residential stability of the common people and improvement of housing welfare. This study conducted an empirical analysis through a proportional odds logit model for households using Bogeumjari Loan, a policy mortgage for the housing vulnerable class, based on the Korea Housing Finance Corporation's '2021 Housing Finance and Demand Survey'. As a result of the analysis, the factors that affect Bogeumjari Loan user satisfaction are loan income requirements, loan maturity, increase in existing loan limit, educational level, type of product used, number of children, comparison with other products before joining, and finally single householder in order. In order to vitalize policy mortgage, it is considered that flexible operation considering the factors that most affect the satisfaction of policy mortgage users, such as housing price loan limit, income requirement, maturity, and repayment capacity, is necessary. As interest rates have risen sharply recently, the burden of repayment of interest on home mortgage loan users has been increasing, so the role of policy mortgage for the housing vulnerable class is expected to become more important. In addition, the results of this study can be used as policy data to improve housing welfare for homeless people or low-income consumers who want to buy their own home.

Determinants of Households’ Intention to Take Out or Convert to a Trust-Type Home Pension: Evidence from South Korea

ABSTRACT This study analyzes the determinants of Korean households’ intention to take out a trust-type home pension (Korean reverse mortgage) or switch from a collateral-type home pension to a trust-type home pension. Using a proportional odds logit model, we analyzed 2018 survey data on home pension demand. Trust-type home pension enrollment is most significantly influenced by homeowners’ relationship with children after enrollment, followed by homeowners’ need for retirement-related financial education. The presence (or absence) of a homeowner’s spouse most significantly influenced conversion intention, affirming that for homeowners with a spouse, their priority in pension selection was securing their spouse’s future entitlement.

신탁방식 주택연금 전환의향 결정요인에 관한 연구

This study analyzed the factors of conversion intention of households using the proportional odds logit model based on the 2018 Reverse Mortgage Demand Survey to seek revitalization of the trust-type reverse mortgage system. As a result of the analysis, the main factors in the conversion to the trust-type were the presence or absence of a spouse, education level, child support, degree of recommendation, sufficient retirement life of the assets held, adult child opinion, and level of satisfaction. It was also confirmed that households using reverse mortgage have a high understanding and acceptability of the trust-type reverse mortgage system and the need to prepare additional retirement income. In conclusion, the trust-type reverse mortgage is expected to be a very useful means not only for new subscribers but also for existing housing pension subscribers. Therefore, it is important to improve related infrastructure so that existing housing pension subscribers can switch to the trust method. At the same time, efforts to improve the system and develop linked products of trust-type reverse mortgage need to be continued.

Do educational inequalities between marital partners influence the empowerment of educated women in ECOWAS?

In an environment where the question of women's empowerment is a burning issue and where inequalities against women in education persist even though they are increasingly educated, it made sense to study the effect of educational inequalities between marital partners on women's empowerment. Thus, using the latest rounds of the Demographic and Health Survey (DHS) women's file from the four major West African economies including Nigeria (2013), Cote d'Ivoire (2011-12), Ghana (2014) and Senegal (2017), and then using the generalized ordered partial proportional odds logit model, we arrived at the result that, inequalities in education between marital partners increase the likelihood of women belonging to the lower empowerment categories. Thus, the policies put in place by States to reduce educational inequalities between men and women must not only make it possible to increase the number of educated women, but above all, these policies must emphasise the achievement of a higher education level by the said women. This is mainly with a view to improving women's ability to participate in decision-making and collective action, allowing them to gain access to sufficiently paid employment to support themselves, and to increasing their ability to experience a violence-free life by having recourse to formal institutions and to confront traditional or social norms that are harmful to them.

The association between food insecurity and mental health during the COVID-19 pandemic

ObjectiveTo explore the association between food insecurity and mental health outcomes among low-income Americans during the COVID-19 pandemic.MethodsWe conducted a survey of 2714 low-income respondents nationwide from June 29, 2020 to July 21, 2020. A proportional odds logit model was employed to estimate the associations between food insecurity and anxiety and between food insecurity and depression.ResultsFood insecurity is associated with a 257% higher risk of anxiety and a 253% higher risk of depression. Losing a job during the pandemic is associated with a 32% increase in risk for anxiety and a 27% increase in risk for depression.ConclusionsFood insecurity caused by the pandemic was associated with increased risk of mental illness. The relative risk of mental illness from being food insecure is almost three-fold that of losing a job during the pandemic. Public health measures should focus on getting direct subsidies of food purchases to poor families, especially families with children. They should also reduce the stigma and shame that is associated with accepting charitable foods.

Investigating factors affecting injury severity in bicycle–vehicle crashes: a day-of-week analysis with partial proportional odds logit models

Cyclists are vulnerable road users and prone to experience severe injury when accidents occur. Both driving and riding behaviors of drivers and cyclists could vary in different days of week, which further influences the injury severity of crashes involving cyclists. This study aims to investigate the factors that affect injury severity in crashes involving cyclists on weekdays and weekends separately using police-reported data ranging from 2007 to 2018 in North Carolina. The impact of cyclist, driver, vehicle, road, environment, and crash characteristics on injury severities are explored. Ordered logit model and partial proportional odds logit models are developed respectively for the injury severities of crashes on weekdays and weekends. Different sets of significant factors are identified for weekdays and weekends. For the common factors identified for both weekdays and weekends, the influencing extent of significant variables varies significantly between weekdays and weekends. Older-aged cyclists, riding direction, pickup, older-aged drivers, male drivers and periods of 0–5:59 and 10–14:59 are only found significant on weekdays while speed limits of 45–55 mph, piedmont areas, commercial development, head-on, and non-roadway locations are only found significant for injury severities of crashes on weekends. Speed limits, time of day, alcohol usage are found to have different or even opposite impacts on the injury severities of crashes on weekdays and weekends.

Pedestrian Injury Severities in Pedestrian-Vehicle Crashes and the Partial Proportional Odds Logit Model: Accounting for Age Difference

This study investigates factors that significantly contribute to the severity of pedestrian injuries resulting from pedestrian-vehicle crashes. Multinomial logit (MNL) models, mixed logit (ML) models, and ordered logit/probit models have been widely used in modeling crash injury severity, including pedestrian injury severity in pedestrian-vehicle crashes. However, both MNL and ML models treat injury severity levels as non-ordered, ignoring the inherent hierarchical nature of crash injury severities, and the data used in ordered logit models need to be strictly subjected to the proportional odds (PO) assumption. In this study, a partial proportional odds (PPO) logit model approach is employed to explore the issues of pedestrian safety associated with each age group: young (aged under 24), middle-aged (aged 25–55), and older pedestrians (aged over 55). Data used in this study are police-reported pedestrian crash data collected from 2007 to 2014 in North Carolina. A variety of motorist, pedestrian, environmental, and roadway characteristics are inspected. Results from likelihood ratio tests statistically show the better performance of developing separate injury severity models for each age group compared with estimating a single model utilizing all data. Relevant parameter estimates and associated marginal effects are used to interpret the results, followed by recommendations made in the concluding section.

Modelling the severity of pedestrian injury in pedestrian—vehicle crashes in North Carolina: A partial proportional odds logit model approach

More than 2,000 pedestrians are involved in reported traffic crashes with vehicles in North Carolina every year. Of these, 10% to 20% of these pedestrians were killed or severely injured. Studies are needed to explore the reasons under such situation for improvement. Traditional methods for modeling crash injury severities use multinomial logit (MNL) models, mixed logit (ML) models, or ordered logit/probit models. However, traditional MNL and ML models treat injury severity levels as nonordered, ignoring the inherent hierarchical nature of crash injury severities, whereas the data used in ordered logit models should be strictly subjected to the proportional odds (PO) assumption. A partial proportional odds (PPO) logit model approach is applied here to address these concerns. The predictors in the PPO model can have different effects on different levels of the dependent variable who violates the PO assumption. This study uses police-reported pedestrian crash data collected from 2007 to 2014 in North Carolina. A variety of motorist, pedestrian, environmental, roadway characteristics are examined. Comparisons of different models are also made and results show that PPO outperforms others. Parameter estimates and associated marginal effects are calculated and used to interpret the model and evaluate the significance of each dependent variables, followed by recommendations in the Conclusion.

Adjusting for COPD severity in database research: developing and validating an algorithm.

PurposeWhen comparing chronic obstructive lung disease (COPD) interventions in database research, it is important to adjust for severity. Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines grade severity according to lung function. Most databases lack data on lung function. Previous database research has approximated COPD severity using demographics and healthcare utilization. This study aims to derive an algorithm for COPD severity using baseline data from a large respiratory trial (UPLIFT).MethodsPartial proportional odds logit models were developed for probabilities of being in GOLD stages II, III and IV. Concordance between predicted and observed stage was assessed using kappa-statistics. Models were estimated in a random selection of 2/3 of patients and validated in the remainder. The analysis was repeated in a subsample with a balanced distribution across severity stages. Univariate associations of COPD severity with the covariates were tested as well.ResultsMore severe COPD was associated with being male and younger, having quit smoking, lower BMI, osteoporosis, hospitalizations, using certain medications, and oxygen. After adjusting for these variables, co-morbidities, previous healthcare resource use (eg, emergency room, hospitalizations) and inhaled corticosteroids, xanthines, or mucolytics were no longer independently associated with COPD severity, although they were in univariate tests. The concordance was poor (kappa = 0.151) and only slightly better in the balanced sample (kappa = 0.215).ConclusionCOPD severity cannot be reliably predicted from demographics and healthcare use. This limitation should be considered when interpreting findings from database studies, and additional research should explore other methods to account for COPD severity.

Effect Displays inRfor Multinomial and Proportional-Odds Logit Models: Extensions to theeffectsPackage

Based on recent work by Fox and Andersen (2006), this paper describes substantial extensions to the effects package for R to construct effect displays for multinomial and proportional-odds logit models. The package previously was limited to linear and generalized linear models. Effect displays are tabular and graphical representations of terms — typically high-order terms — in a statistical model. For polytomous logit models, effect displays depict fitted category probabilities under the model, and can include point-wise confidence envelopes for the effects. The construction of effect displays by functions in the effects package is essentially automatic. The package provides several kinds of displays for polytomous logit models.

Effect Displays for Multinomial and Proportional-Odds Logit Models

An “effect display” is a graphical or tabular summary of a statistical model based on high-order terms in the model. Effect displays have previously been defined by Fox (1987, 2003) for generalized linear models (including linear models). Such displays are especially compelling for complicated models—for example, those including interactions or polynomial terms. This paper extends effect displays to models commonly used for polytomous categorical response variables: the multinomial logit model and the proportional-odds logit model. Determining point estimates of effects for these models is a straightforward extension of results for the generalized linear model. Estimating sampling variation for effects on the probability scale in the multinomial and proportional-odds logit models is more challenging, however, and we use the delta method to derive approximate standard errors. Finally, we provide software for effect displays in the R statistical computing environment.