Latent Class Multinomial Logit Research Articles

Analyzing driver injury severity in two-vehicle rear-end crashes considering leading-following configurations based on passenger car and light truck involvement

Rear-end crash is a major type of traffic crashes leading to a large number of injuries and fatalities each year, and passenger cars and light trucks are two main vehicle types in rear-end crashes on US roadways. Passenger cars and light trucks are different in size, vehicle mass and driver’s vision. It is necessary to investigate the driver injury outcome patterns in rear-end crashes between passenger cars and light trucks considering crash configurations regarding the leading and following vehicle types. This study employs latent class multinomial logit (MNL) model to examine the risk factors on driver injury severity along with heterogeneity in variable effects presented by the cluster pattern in two-vehicle rear-end crashes involving passenger cars and light trucks, considering four crash configuration types, i.e., a passenger car struck by a passenger car, a light truck struck by a light truck, a passenger car struck by a light truck, and a light truck struck by a passenger car as exploratory variables. A model with two latent classes, which indicates the heterogeneity in variable effects among all the observations, is found to best fit the 7-year crash dataset from Washington State. The pseudo-elasticities are calculated to quantify the marginal effects of the contributing factors. The risk factors curve and sloping road condition, driver without seatbelt, and driver age of 65 and above increase driver fatality and serious injury risk greatly, and these three factors contribute from different latent classes. The crash configuration of a passenger car struck by a light truck is found to be one of class characteristics factors, which indicates that the heterogeneity exists between these two vehicle types. This factor is also a risk factor of injury. Furthermore, the leading vehicle is found to be much more vulnerable and closely related to injury, especially when it is in the crash of a passenger car struck by a light truck. The latent classes discovered give theoretical evidence of how to appropriately select subset data for further model construction for practical interest of serious injury prevention. The risk factors and their influence on injury severity provide beneficial insights on developing relevant countermeasures and strategies for injury severity mitigation on rear-end crashes involving passenger cars and light trucks.

Temporal assessment of injury severities of two types of pedestrian-vehicle crashes using unobserved-heterogeneity models

This study explores the temporal instability and non-transferability of the determinants affecting injury severities of pedestrians struck by motorcycles and non-motorcycles. Using the pedestrian-vehicle crash data in Rawalpindi, Pakistan, over three years (2017–2019), three possible crash injury severity categories (minor injury, severe injury, and fatal injury) are estimated using alternative models to account for unobserved heterogeneity. These are a random-parameters multinomial logit (RP-ML) model with heterogeneity in means and variances and a latent-class multinomial logit (LC-ML) model with class probability functions. Temporal instability and non-transferability in the effects of explanatory variables are confirmed using a series of likelihood ratio tests based on the two alternative models. Various variables are observed to determine pedestrian-injury severities, and the estimation results show significant temporal instability and non-transferability in both RP-ML and LC-ML models. However, several explanatory variables produce relatively temporally stable and transferable effects, providing valuable insights to implement effective countermeasures from a long-term perspective. Moreover, out-of-sample predictions are simulated to confirm the temporal instability and non-transferability. At the same time, the LC-ML models produce higher differences for temporal instability and lower differences for non-transferability compared to the RP-ML model. Understanding and depth comparing the estimation results, likelihood ratio tests, and out-of-sample predictions using alternative models is a promising direction for future research to explore how the observed and unobserved heterogeneity can be estimated in terms of temporal instability and non-transferability.

Is it more important to increase carbon sequestration, biodiversity, or jobs? A case study of citizens' preferences for forest policy in Finland

Forest management plays an important role in mitigating the most serious threats to the well-being of current and future generations: climate change and biodiversity loss. We examine citizens' preferences for alternative forest management policy goals, including the provision of carbon sinks, biodiversity, and forest sector jobs, using a discrete choice experiment in Finland. We explore preference heterogeneity using latent class multinomial logit models finding the highest willingness to pay for biodiversity-oriented forest policy. Comparing three forest policy scenarios, focusing on “Biodiversity,” “Climate,” or “Bioeconomy,” citizens preferred the overall environmental goals over economic ones. We observe considerable heterogeneity in respondents' preferences identifying five latent preference classes. The results provide valuable background information on citizens' values and preferences on tradeoffs built into future revisions of forest policy.

Alternative unobserved heterogeneity models to analyze injury severity of expressway crashes in different tunnel types

This study examines the non-transferability of the determinants affecting injury severities of tunnels of three grades classified by different lengths. Based on tunnel crash data in Shanxi Province during (2012–2018), three possible crash injury severity categories (no injury, minor injury, and severe injury) are estimated. Two alternate models were used to account for unobserved heterogeneity: a random parameters multinomial logit (RPML) model with heterogeneity in means and variances and a latent-class multinomial logit (LCML) model with class probability functions. Non-transferability in the effects of explanatory variables is confirmed using two series of likelihood ratio tests based on the two alternate models. A wide variety of variables is observed to determine pedestrian-injury severities. The estimation results show significant non-transferability across the three grades in both RPML and LCML models. In addition, several explanatory variables produce relatively transferable effects, which can provide some valuable insights to implement effective countermeasures. Moreover, out-of-sample predictions are simulated to confirm the non-transferability, while the LC models produce higher differences for the non-transferability than RP models. According to estimation results, several practical measurements concerning education programs, traffic enforcement, emergency response systems, and traffic safety measurements are implemented to eliminate tunnel crash outcomes. Understanding and depth comparing the estimation results, likelihood ratio tests and out-of-sample predictions using alternate models is a promising direction for future research to explore how the observed and unobserved heterogeneity can be estimated.

A latent class multinomial logit analysis of factors associated with pedestrian injury severity of inter-urban highway crashes

Over the years, the uncontrolled interaction of human and high-speed vehicular activities at settlement areas along highways in Ghana has resulted in many pedestrian fatalities and injuries. This phenomenon has been attributed to the land-use and right-of-way planning and lack of pedestrian crossing facilities for safe crossing of highways. The slow response to developing strategies to reduce pedestrian fatalities along the nation’s highways has led to many public protests. To advance a data-driven and evidence-based approach to finding appropriate countermeasures, this study investigated the factors associated with pedestrian injury outcomes of inter-urban highway crashes in Ghana. Latent class multinomial logit modeling method was employed to account for unobserved heterogeneity in a five-year pedestrian-vehicle crash data recorded on highways in Ghana. The model estimation results show that speeding, hit and run and crashes that involve buses were more likely to result in fatal injury while crashes that occurred at highway sections with no shoulder were more likely to result in hospitalized injury. The findings of the study provide basis for the development of appropriate countermeasures to reduce the number of pedestrian deaths and injuries on high-speed inter-urban highways in Ghana and other countries with similar characteristics in the sub-region.

Investigating the preferences between shared and non-shared ride-hailing services across user groups

We investigate the effects of travel duration and trip costs on mode choice between shared (e.g., Uber) and non-shared (e.g., the former Uber Pool) ride-hailing services. Using the City of Chicago as a case study, we employ a novel method by linking individual characteristics from the Chicago region’s household travel survey to trip characteristics from the City of Chicago’s ride-hailing trip data between November 2018 and December 2019. The findings from a binomial logit analysis and a latent class multinomial logit analysis indicate that travel duration and trip costs might not be the determinants for mode choice between shared and non-shared ride-hailing services. Under the current fare structure, some passengers might prefer one mode over the other regardless of the cost and time difference between the two types of services, while some might consider the differences in time and cost between the two modes not significant enough for switching mode. Our method allows us to investigate mode choice for the shared and non-shared ride-hailing services without relying on skim tables from regional travel models or stated preference surveys.

Patient Preferences for Multi-Cancer Early Detection (MCED) Screening Tests.

Emerging blood-based multi-cancer early detection (MCED) tests can detect a variety of cancer types across stages with a range of sensitivity, specificity, and ability to predict the origin of the cancer signal. However, little is known about the general US population's preferences for MCED tests. To quantify preferences for MCED tests among US adults aged 50-80 years using a discrete choice experiment (DCE). To quantify preferences for attributes of blood-based MCED tests, an online DCE was conducted with five attributes (true positives, false negatives, false positives, likelihood of the cancer type unknown, number of cancer types detected), among the US population aged 50-80 years recruited via online panels and social media. Data were analyzed using latent class multinomial logit models and relative attribute importance was obtained. Participants (N = 1700) were 54% female, mean age 63.3 years. Latent class modeling identified three classes with distinct preferences for MCED tests. The rank order of attribute importance based on relative attribute importance varied by latent class, but across all latent classes, participants preferred higher accuracy (fewer false negatives and false positives, more true positives) and screenings that detected more cancer types and had a lower likelihood of cancer type unknown. Overall, 72% of participants preferred to receive an MCED test in addition to currently recommended cancer screenings. While there is significant heterogeneity in cancer screening preferences, the majority of participants preferred MCED screening and the accuracy of these tests is important. While the majority of participants preferred adding an MCED test to complement current cancer screenings, the latent class analyses identified a small (16%) and specific subset of individuals who value attributes differently, with particular concern regarding false-negative and false-positive test results, who are significantly less likely to opt-in.

Purchase preferences-based air passenger choice behavior analysis from sales transaction data

Travel providers such as airlines are becoming more and more interested in understanding how passengers choose among alternative products, especially the purchasing preferences of passengers. Getting information of air passenger choice behavior helps them better display and adapt their offer. Discrete choice models are appealing for airline revenue management (RM). In this paper, we apply latent class multinomial logit model (LC-MNL) to passenger choice behavior. The analysis based on actual sales transaction data reveals the purchase preferences of different passenger types. According to the distribution of the market, we divide passengers into three groups: low-price oriented, high-price oriented and no specific price preference. The low-price oriented passengers only choose products from the set which consists of the lowest price cabin classes of each flight, while the high-price oriented passengers do the opposite. Considering that the passenger types in the transaction sales data are unknown, the latent class passenger choice model can better represent their heterogeneous purchasing preference. An improved EM algorithm is applied to solve the LC-MNL. In the improved EM algorithm, an indicator function containing both the type of passengers and the first choice information in period t is devised, the iterative process of the EM algorithm is more effective consequently. The proposed model and algorithm are evaluated on actual aviation sales transaction data in China. Experimental results show that the passenger choice behavior analysis based on the specific purchasing preferences performs well on actual aviation sales transaction data.

Examining the who, what, and how of risky driving related crashes in residential areas

This study contributes to efforts towards creating livable communities, by identifying and proposing countermeasures for factors that are associated with residential area crashes. Three risky driving pre-crash actions: driving under the influence of alcohol or drug, aggressive driving, and speeding were considered. The study used residential area crash data from Alabama, USA and developed a latent class multinomial logit model to identify factors associated with the pre-crash actions. The results showed that male drivers had 0.44% and 2.09% higher chances to be at fault in driving under the influence of alcohol or drug and aggressive driving crashes, respectively, while female drivers were found to have a 1.82% higher likelihood to be at fault in speeding-related crashes but less likely to be at fault in aggressive driving or driving under the influence of alcohol or drug-related crashes in residential areas. Speeding and driving under the influence of were found to be the leading contributing factors in rural area residential crashes whereas aggressive driving was more pronounced in urban areas. These findings provide a basis for the implementation of appropriate crash countermeasures to reduce residential area crashes, as part of efforts geared towards promoting community safety and livability across the state.

A temporal assessment of distracted driving injury severities using alternate unobserved-heterogeneity modeling approaches

This study explores temporal shifts in the effects of explanatory variables on the injury severity outcomes of crashes involving distracted driving. Using data from distracted driving crashes on Kansas State highways over a four-year period (from 2014 to 2017 inclusive), separate yearly models of driver-injury severities (with possible outcomes of severe injury, minor injury, and no injury) were estimated using two alternate modeling approaches to account for possible unobserved heterogeneity: a latent-class multinomial logit with class probability functions and a random parameters logit with possible heterogeneity in the means and variances of random parameters. Likelihood ratio tests were conducted to determine if model parameter estimates have shifted over time. A wide range of variables were found to statistically influence driver-injury severities and the findings show that were statistically significant temporal shifts in parameter estimates in both the random parameters and latent class modeling approaches. These shifts are likely the result of changes in driver behavior, improvements in vehicle and highway safety features, changes in communication technologies, and other temporally shifting trends. However, while out-of-sample simulations show that the two modeling approaches both indicate that distracted driving crashes have become less severe over time, the alternate approaches produced substantially different injury-severity predictions, suggesting the need for future research to explore how unobserved heterogeneity can best be modeled in temporal contexts.

Patterns of SARS-CoV-2 Testing Preferences in a National Cohort in the United States: Latent Class Analysis of a Discrete Choice Experiment.

BackgroundInadequate screening and diagnostic testing in the United States throughout the first several months of the COVID-19 pandemic led to undetected cases transmitting disease in the community and an underestimation of cases. Though testing supply has increased, maintaining testing uptake remains a public health priority in the efforts to control community transmission considering the availability of vaccinations and threats from variants.ObjectiveThis study aimed to identify patterns of preferences for SARS-CoV-2 screening and diagnostic testing prior to widespread vaccine availability and uptake.MethodsWe conducted a discrete choice experiment (DCE) among participants in the national, prospective CHASING COVID (Communities, Households, and SARS-CoV-2 Epidemiology) Cohort Study from July 30 to September 8, 2020. The DCE elicited preferences for SARS-CoV-2 test type, specimen type, testing venue, and result turnaround time. We used latent class multinomial logit to identify distinct patterns of preferences related to testing as measured by attribute-level part-worth utilities and conducted a simulation based on the utility estimates to predict testing uptake if additional testing scenarios were offered.ResultsOf the 5098 invited cohort participants, 4793 (94.0%) completed the DCE. Five distinct patterns of SARS-CoV-2 testing emerged. Noninvasive home testers (n=920, 19.2% of participants) were most influenced by specimen type and favored less invasive specimen collection methods, with saliva being most preferred; this group was the least likely to opt out of testing. Fast-track testers (n=1235, 25.8%) were most influenced by result turnaround time and favored immediate and same-day turnaround time. Among dual testers (n=889, 18.5%), test type was the most important attribute, and preference was given to both antibody and viral tests. Noninvasive dual testers (n=1578, 32.9%) were most strongly influenced by specimen type and test type, preferring saliva and cheek swab specimens and both antibody and viral tests. Among hesitant home testers (n=171, 3.6%), the venue was the most important attribute; notably, this group was the most likely to opt out of testing. In addition to variability in preferences for testing features, heterogeneity was observed in the distribution of certain demographic characteristics (age, race/ethnicity, education, and employment), history of SARS-CoV-2 testing, COVID-19 diagnosis, and concern about the pandemic. Simulation models predicted that testing uptake would increase from 81.6% (with a status quo scenario of polymerase chain reaction by nasal swab in a provider’s office and a turnaround time of several days) to 98.1% by offering additional scenarios using less invasive specimens, both viral and antibody tests from a single specimen, faster turnaround time, and at-home testing.ConclusionsWe identified substantial differences in preferences for SARS-CoV-2 testing and found that offering additional testing options would likely increase testing uptake in line with public health goals. Additional studies may be warranted to understand if preferences for testing have changed since the availability and widespread uptake of vaccines.

A Multi-Country Comparison of Consumers’ Preferences for Imported Fruits and Vegetables

Within Asia, imported fruits and vegetables are often considered as a delicacy and of high value, and are increasingly demanded compared to local products. There are numerous significant factors involved with consumers’ characteristics and their corresponding values towards these products. This study investigates potential consumers and their preferences towards imported fruits and vegetables in three Asian countries: Taiwan, Japan, and Indonesia. A total of 1350 survey responses collected from Taiwan, Japan, and Indonesia are examined by a best–worst scaling method with a latent class multinomial logit model. Results show that consumers tend to choose imported fruits that are not commonly provided by domestic producers. While a food safety certified label and freshness are consistently identified as the most and second most important food values for Taiwanese, Japanese, and Indonesian consumers, price is still an important factor for certain consumer groups. The majority of Taiwanese and Japanese consumers (i.e., female, higher education, and from an urban area) prefer imported fruits and vegetables, while the majority of Indonesian consumers do not pay much attention to imported fruits and vegetables. While Taiwan, Japan, and Indonesia are island countries, the novelty of this study shows that consumer preferences do not behave the same. The implications of this study should be of interest to producers and exporters who wish to positively impact the design of their international marketing strategies.

A Comparative Empirical Study of Discrete Choice Models in Retail Operations

Choice-based demand estimation is a fundamental task in retail operations and revenue management, providing necessary input data for inventory control, assortment, and price-optimization models. The task is particularly difficult in operational contexts where product availability varies over time and customers may substitute into the available options. In addition to the classical multinomial logit (MNL) model and extensions (e.g., nested logit, mixed logit, and latent-class MNL), new demand models have been proposed (e.g., the Markov chain model), and others have been recently revisited (e.g., the rank list-based and exponomial models). At the same time, new computational approaches were developed to ease the estimation function (e.g., column-generation and expectation-maximization (EM) algorithms). In this paper, we conduct a systematic, empirical study of different choice-based demand models and estimation algorithms, including both maximum-likelihood and least-squares criteria. Through an exhaustive set of numerical experiments on synthetic, semisynthetic, and real data, we provide comparative statistics of the predictive power and derived revenue performance of an ample collection of choice models and characterize operational environments suitable for different model/estimation implementations. We also provide a survey of all the discrete choice models evaluated and share all our estimation codes and data sets as part of the online appendix. This paper was accepted by Vishal Gaur, operations management.

The Latent Class Multinomial Logit Model for Modeling Front-Seat Passenger Seatbelt Choice, Considering Seatbelt Status of Driver

The literature review highlighted the impacts of drivers’ behavior on passengers’ attitudes in the choice of seatbelt usage. However, limited studies have been done to determine those impacts. Studying the passengers’ seatbelt use is especially needed to find out why passengers choose not to buckle up, and consequently it helps decision makers to target appropriate groups. So, this study was conducted to find drivers’ characteristics that might impact the passenger’s seatbelt use, in addition to other passengers’ characteristics themselves. While performing any analysis, it is important to use a right statistical model to achieve a less biased point estimate of the model parameters. The latent class multinomial logit model (LC-MNL) can be seen as an alternative to the mixed logit model, replacing the continuous with a discrete distribution, by capturing possible heterogeneity through membership in various clusters. In this study, instead of a response to the survey or crash observations, we employed a real-life observational data for the analysis. Results derived from the analysis reveal a clear indication of heterogeneity across individuals for almost all parameters. Various socio-demographic variables for class allocation and models with different latent numbers were considered and checked in terms of goodness of fit. The results indicated that a class membership with three factors based on vehicle type would result in a best fit. The results also highlighted the significant impacts of driver seatbelt status, time of a day, distance of traveling, vehicle type, and driver gender, instead of passenger gender, as some of the factors impacting the passengers’ choice of seatbelt usage. In addition, it was found that the belting status of passengers is positively associated with the belting condition of drivers, highlighting the psychological behavioral impact of drivers on passengers. Extensive discussion has been made regarding the implications of the findings.

Development of Population Tariffs for the ICECAP-A Instrument for Hungary and their Comparison With the UK Tariffs

ObjectivesThe ICEpop CAPability measure for Adults (ICECAP-A) was developed to assess the capability well-being of adults for use in economic evaluations. Currently, ICECAP-A tariffs are available only for the UK population. The objectives of this study were to develop a Hungarian tariff set for the ICECAP-A instrument and to explore intercountry differences between the Hungarian and the UK value sets. MethodsA survey was conducted by computer-assisted personal interviews on a sample representative of the Hungarian adult population (N = 1000) to elicit their preferences regarding ICECAP-A attributes with the use of a best-worst scaling choice task. A latent class multinomial logit model with continuous variance scale was used to estimate the weights for each of the 4 capability levels of all 5 ICECAP-A attributes, namely, attachment, stability, achievement, enjoyment, and autonomy. ResultsThe model identified 2 preference classes with approximately equal share. The first class had a stronger relative preference for autonomy and achievement, whereas the second class had a strong preference for attachment. Multivariate analysis of the classes revealed that women, pensioners, people who are married or living in a partnership, and people with poorer health status are characteristics associated with the latter class membership (preference for attachment). Population tariffs were estimated from the model. Overall, attachment was found to be the most important attribute, followed by stability, enjoyment, achievement, and autonomy. ConclusionsHungarian tariffs are largely consistent with those found for the United Kingdom; nevertheless, autonomy seems to be less important in Hungary compared with the United Kingdom.

An Optimal Greedy Heuristic with Minimal Learning Regret for the Markov Chain Choice Model

We study the assortment optimization problem and show that local optima are global optima for all discrete choice models that can be represented by the Markov Chain model. We develop a forward greedy heuristic that finds an optimal assortment for the Markov Chain model and runs in $O(n^2)$ iterations. The heuristic has performance bound $1/n$ for any regular choice model which is best possible among polynomial heuristics. We also propose a backward greedy heuristic that is optimal for Markov chain model and requires fewer iterations. Numerical results show that our heuristics performs significantly better than the estimate then optimize method and the revenue-ordered assortment heuristic when the ground truth is a latent class multinomial logit choice model. Based on the greedy heuristics, we develop a learning algorithm that enjoys asymptotic optimal regret for the Markov chain choice model and avoids parameter estimations, focusing instead on binary comparisons of revenues.

Heterogeneity in physical activity participation of older adults: A latent class analysis

Participation in leisure and transport-related physical activities is generally viewed as the main means of increasing overall physical activity levels of older adults. The success of any intervention strategy depends on how well the strategy fits older people's physical activity patterns. This study therefore examines heterogeneity in physical activity patterns of older adults and the relationships between these patterns and socio-demographic and neighborhood environmental characteristics. This is done using a latent class multinomial logit model based on one-week diary data collected on 363 Dalian Chinese older adults aged 60 and over in 2017. The results show that two segments of physical activity patterns of older adults can be identified: leisure physical activities oriented and both transport-related and leisure physical activities oriented. Age, physical limitation and having grandchildren in the household have an important influence on physical activity patterns of older people. However, there are no significant relationships between physical activity patterns and various environmental characteristics, except accessibility of local shops, condition of footpaths and safety from crime. The empirical findings have implications for the development of tailored interventions to maintain or promote physical activity of older adults.

Comparison of food values for consumers’ preferences on imported fruits and vegetables within Japan, Taiwan, and Indonesia

International agricultural trade in fruits and vegetables has received much attention as an effective method to create a sustainable agriculture industry for a country. This study focuses on consumers’ perceptions of food values toward imported fruits and vegetables in Japan, Taiwan, and Indonesia by using the Best-Worst Scaling method. Seven food values were examined in this study: labeling product origin, food safety certification, high quality appearance, domestic rarity, price, how the product was grown, and freshness. An online survey was conducted and 1,350 total valid respondents were collected (500 Japanese, 333 Taiwanese, and 517 Indonesian). The Latent Class Multinomial Logit model was used to analyze consumers for each country. Results revealed that food safety certification and freshness were found as the most important and second most important food values for the majority group of consumers in each of these three countries, respectively. However, the remainder of food values do not have the same importance for each country. The study should help governments and international marketers enhance their agricultural trade policies and marketing strategies for these targeted markets.

Role of Passengers in Single-Vehicle Drunk-Driving Crashes: An Injury-Severity Analysis

Background: Drunk-driving is a major crash risk factor, and crashes resulting from this risky behavior tend to be serious and have significant economic and societal impacts. The presence of passengers and their demographics and activities can influence risky driving behaviors such as drunk-driving. However, passengers could either be an “enabling” factor to take more risks or could be an “inhibiting” factor by ensuring safe driving by a drunk-driver. Objective: This study examines whether the presence of passengers affects the contributing factors of single-vehicle (SV) drunk-driving crashes, by presenting a severity analysis of single- and multi-occupant SV drunk-driving crashes, to identify risk factors that contribute to crash severity outcomes, for the effective implementation of relevant countermeasures. Method: A total of 7407 observations for 2012–2016 from the crash database of the State of Alabama was used for this study. The variables were divided into six classes: temporal, locational, driver, vehicle, roadway, and crash characteristics and injury severities into three: severe, minor, and no injury. Two latent class multinomial logit models—one each for single- and multi-occupant crashes—were developed, to analyze the effects of significant factors on injury severity outcomes using marginal effects. Results: The estimated results show that collision with a ditch, run-off road, intersection, winter season, wet roadway, and interstate decreased the probability of severe injuries in both single- and multi-occupant crashes, whereas rural area, road with downward grade, dark and unlit roadway, unemployed driver, and driver with invalid license increased the likelihood of severe injuries for both single- and multi-occupant crashes. Female drivers were more likely to be severely injured in single-occupant crashes, but less likely in multi-occupant crashes. A significant association was found between severe injuries and weekends, residential areas, and crash location close (<25 mi ≈40.23 km) to the residence of the at-fault driver in multi-occupant crashes. Sport utility vehicles were found to be safer when driving with passengers. Conclusions: The model findings show that, although many correlates are consistent between the single- and multi-occupant SV crashes that are associated with locational, roadway, vehicle, temporal, and driver characteristics, their effect can vary across the single- and multi-occupant driving population. The findings from this study can help in targeting interventions, developing countermeasures, and educating passengers to reduce drunk-driving crashes and consequent injuries. Such integrated efforts combined with engineering and emergency response may contribute in developing a true safe systems approach.

(Rescaled) Multi‐Attempt Approximation of Choice Model and Its Application to Assortment Optimization

When designing an assortment, a fundamental trade‐off exists between estimating consumer behavior more precisely and solving the corresponding optimization problem efficiently. For instance, a mixture of logit model can closely approximate any random‐utility choice model; however, it is not suitable for optimization when the number of customer segments and the number of products considered are large. As many companies are using big data and marketing analytics toward microsegmenting consumers, a choice model that is amenable in the problem size becomes necessary for assortment design. In this work, we provide a new approach to approximate any random‐utility choice model by characterizing the estimation errors in a classical exogenous demand model and significantly improving its performance with a rescaling method. We show that the resulting approximation is exact for Multinomial Logit (MNL). If, however, the underlying true choice model is not MNL, we show numerically that the approximation under our so‐called rescaled two‐attempt model outperforms the widely used MNL approximation, and provides performance close to the Markov chain approximation (in some cases, it performs better than the Markov chain approximation). Our proposed approximation can be used to solve a general assortment optimization problem with a variety of (linear) real‐world constraints. In contrast to the more direct Mixed Integer Optimization (MIO) approach that utilizes Latent Class Multinomial Logit (LC‐MNL), whose running time increases exponentially in the number of mixtures, our approximation yields an alternative MIO formulation whose empirical running time is independent of the number of mixtures.