Multimodal preference heterogeneity in choice-based conjoint analysis: a simulation study

Screening for interactions between design factors and demographics in choice-based conjoint

Sensitivity evaluation of machine learning-based calibrated transportation mode choice models: A case study of Alexandria City, Egypt

This paper investigates freight market segmentation to seek the potential market of modal shift from road to rail-based intermodal (only pre- and end- haulage are transported by road), using discrete choice analysis. The Defficient design approach was used to generate SP scenarios. The data, including 2160 SP observations of 135 companies, was collected in China's Yangtze River Delta region. The multinomial logit (MNL) model, the mixed logit (ML) model, and the latent class (LC) model were constructed successively. Freight mode attributes (cost, time and reliability), shipment characteristics (distance, time threshold), cargo characteristics (value and density) were considered. In addition, the logistics elements such as the stages of production process and psychological characteristics such as choice inertia were also introduced in this paper. The results show that the LC model with segment variables performs better than the counterpart without segment variables, the MNL model and the ML model. Using the LC model, shippers are classified into four market segments. There are significant differences in shippers' preference among different classes. Based on the class membership function, the characteristic of each class can be identified by a combination of different attributes. The difficulty of achieving the modal shift from road to rail-based intermodal is ranked according to the values of attributes, further determining the priority of the potential market. This study helps to understand the shippers' freight mode choice better. Rail service suppliers can design transportation products that match the shippers' demand characteristics for different classes to improve the market share.

The rapid growth of quick commerce (q-commerce) is changing urban grocery retailing by guaranteeing delivery of essentials, including fresh fruits and vegetables in 10-30 min. However, speed of delivery has implications for postharvest management of highly perishable food items such as leafy greens and climacteric fruits (i.e., bananas, mangoes, tomatoes and peaches). There is a significant lack of research examining how physiological quality preservation methods influence consumer preferences for highly perishable items such as leafy greens and climacteric fruits. This research fills this gap by investigating urban consumers' preferences regarding the trade-off between delivery speed, price and quality preservation on q-commerce platforms. The study was conducted in the Coimbatore district in Tamil Nadu. A structured survey of 110 urban consumers in Coimbatore, Tamil Nadu, was conducted using convenience sampling. Choice-Based Conjoint (CBC) analysis under the Random Utility Theory framework was applied using an orthogonal design with three attributes: delivery speed, quality preservation methods and pricing. Consumer utility values were estimated using the Multinomial logit model (MNL) with significance testing at p<0. 05. The results indicated that delivery speed held the highest relative importance (38 %, p<0.001) in consumer preferences, followed by quality preservation methods (32 %, p<0.001) and price sensitivity (30 %, p<0.001). Exploratory Factor Analysis (EFA) revealed that physiological quality indicators such as freshness, ripeness and shelf life were the most influential latent factor (explaining 37.5 % of variance). Garrett Ranking identified inconsistent ripening (mean = 69.68) and high delivery charges (mean = 56.60) as the top barriers to purchasing fresh produce via quick commerce platforms.

The study addressed the increasing complexity faced by students when selecting a university major at the Universidad de las Fuerzas Armadas – ESPE, considering the expansion of academic offerings and the changing demands of the labor market. Its objective was to identify the key determinants guiding this decision through the application of discrete choice models. The Multinomial Logit and Hierarchical Logit models were employed to examine the influence of personal, institutional, and expectation-related variables, organizing program alternatives into disciplinary categories to capture preference heterogeneity. The results showed that personal factors mainly vocation, individual interests, and long-term professional aspirations exerted the strongest influence on major selection, while institutional and economic considerations played a secondary role. The Hierarchical Logit model exhibited a superior fit compared with the Multinomial Logit model, indicating that disciplinary nesting allowed a more accurate representation of decision patterns. Additionally, expected employability and opportunities for career advancement emerged as critical motivations driving students’ choices. In conclusion, the study demonstrated that discrete choice models provided a robust methodological framework to explain educational decision-making processes, offering valuable evidence to strengthen vocational guidance, align academic programs with market demands, and support institutional planning based on empirical insights.

This paper proposes a new classification method based on Dirichlet process mixtures(DPM) to investigate the classification of the four actions from the action surface EMG(ASEMG) signals. This method first builds a classification model of the data by using the multinomial logit model (MNL). Then a classifier is given by using the classification information of training data. For the features of ASEMG, we use a combined method of the empirical mode decomposition(EMD), Largest Lyapunov exponent and Linear discriminant analysis(LDA) dimension reduction. The highest average classification accuracy rates are over 90%. The results indicate that this classification method could be applied the classification of the ASEMG signals.KeywordsDirichlet process mixturethe action Surface EMG(ASEMG) signalempirical mode decomposition(EMD)Largest Lyapunov exponentLinear discriminant analysis(LDA)

Random parameter logit models address unobserved preference heterogeneity in discrete choice analysis. The latent class logit model assumes a discrete heterogeneity distribution, by combining a conditional logit model of economic choices with a multinomial logit (MNL) for stochastic assignment to classes. Whereas point estimation of latent class logit models is widely applied in practice, stochastic assignment of individuals to classes needs further analysis. In this paper we analyze the statistical behavior of six competing class assignment strategies, namely: maximum prior MNL probabilities, class drawn from prior MNL probabilities, maximum posterior assignment, drawn posterior assignment, conditional individual-specific estimates, and conditional individual estimates combined with the Krinsky–Robb method to account for uncertainty. Using both a Monte Carlo study and two empirical case studies, we show that assigning individuals to classes based on maximum MNL probabilities behaves better than randomly drawn classes in market share predictions. However, randomly drawn classes have higher accuracy in predicted class shares. Finally, class assignment based on individual-level conditional estimates that account for the sampling distribution of the assignment parameters shows superior behavior for a larger number of choice occasions per individual.

Preference heterogeneity for renewable energy technology

BackgroundAntithrombotic drugs are used as preventive treatment in patients with a prior myocardial infarction (MI) in both the acute and chronic phases of the disease. To support patient-centered benefit–risk assessment, it is important to understand the influence of disease stage on patient preferences.ObjectiveThe aim of this study was to examine patient preferences for antithrombotic treatments and whether they differ by MI disease phase.MethodsA discrete-choice experiment was used to elicit preferences of adults in the acute (≤ 365 days before enrolment) or chronic phase (> 365 days before enrolment) of MI for key ischemic events (risk of cardiovascular [CV] death, non-fatal MI, and non-fatal ischemic stroke) and bleeding events (risk of non-fatal intracranial hemorrhage and non-fatal other severe bleeding). Preference data were analyzed using the multinomial logit model. Trade-offs between attributes were calculated as the maximum acceptable increase in the risk of CV death for a decrease in the risk of the other outcomes. To assess the potential effect of sociodemographic and clinical characteristics on patient preferences, subgroups were introduced as interaction terms in logit models.ResultsThe evaluable population included 155 patients with MI in the acute phase of disease and 180 in the chronic phase. The overall population was 82% male, mean age was 64.2 ± 9.6 years, and 93% had not experienced bleeding events or key ischemic events other than MI. Patients valued reduction in the risk of non-fatal intracranial hemorrhage more than CV death (p < 0.01) and CV death more than non-fatal ischemic events (p < 0.01). Preferences were similar in the acute and chronic populations (p = 0.17). However, older patients valued reduction in risk of MI more than younger patients (p = 0.04), and patients with bleeding risk factors valued reduction in the risk of CV death (p = 0.01) and MI (p = 0.01) less than patients without bleeding risk factors. Also, patients who were at high risk of future ischemic events valued reduction of the risk of CV death less than those at low risk (p = 0.01).ConclusionPatient preferences for antithrombotic treatments were unaffected by disease stage but varied by bleeding risk and other factors. This heterogeneity in preferences is an important consideration because it can affect the benefit–risk balance and the acceptability of antithrombotic treatments to patients.Supplementary InformationThe online version contains supplementary material available at 10.1007/s40271-021-00548-6.

The authors integrate concepts in conjoint analysis and discrete choice theory in econometrics to develop a new approach to the design and analysis of controlled consumer choice or resource allocation experiments. The article is concerned with estimating the parameters of conjoint-type functions from discrete choice or allocation data. Emphasis is placed on the multinomial logit model and aggregate choice or allocation data to illustrate the concepts in a series of empirical examples ranging from simple to complex. The authors present limited external validity evidence to support the approach and make comparisons with traditional conjoint approaches.

This paper presents the findings of a study recently conducted in Florida to quantify freight users’ willingness to pay (WTP) for the improvement of transportation-related attributes, particularly reliability. A stated preference survey was developed and administered between January and May 2016. The survey collected responses from 150 shippers, carriers, and forwarders. Econometric models, including mixed and multinomial logit models, were developed to estimate the users’ WTP and to investigate the presence of user heterogeneity. The value of time and the value of reliability were estimated separately for the various user groups. The results indicated that carriers showed the lowest WTP when their WTP was compared with that of other freight users. Shippers without transportation—that is, shippers who contracted out their shipping— exhibited more interest in reducing travel time savings, whereas shippers with transportation showed more sensitivity to reliability. Preference heterogeneity was also explored by commodity group and product type. The results confirmed the findings from past studies and showed significant differences in WTP values when the sources of heterogeneity were considered. This paper contributes to the literature by providing empirical evidence of the quantification of the value of reliability in road freight transportation and the impacts of user heterogeneity. The study results will help advance understanding of the impacts of the performance of transportation systems on the freight industry.

In treating early-onset scoliosis (EOS), pediatric orthopaedic surgeons and families must often choose between growth-friendly interventions such as magnetically controlled growing rods (MCGR) and early definitive posterior spinal fusion (PSF). While each strategy has associated pros and cons, little is understood about the specific preferences of patients' families. Addressing potential nuances in these preferences is crucial for meaningful shared decision-making (SDM). To better understand the factors influencing family decisions, we conducted a choice-based conjoint (CBC) analysis to explore parents' preferences for PSF and MCGR as treatment options. A survey was developed using Sawtooth Software (Lighthouse Studio version 9.2.0) to gather demographic information and preferences on surgical scenarios via CBC analysis. Anonymous participants, recruited via the Prolific crowdsourcing platform, qualified if they were US residents over 18 years of age with children aged 8 to 9. Data analysis involved the Hierarchical Bayes (HB) method to determine the desirability of 5 surgical attributes: curve correction, increase in spinal height, total number of surgeries, duration of treatment, and chance of complications. The study included 432 participants who prioritized the likelihood of complications (relative importance value of 37), followed by total number of surgeries (22), curve correction (21), spine height (12), and treatment duration (8). Female participants and those with daughters assigned significantly higher importance to the treatment's total number of surgeries ( P <0.05). Simulator analysis showed that 86.6% of participants would hypothetically prefer PSF, while only 13.4% would opt for growing rods based on the modeled participant choices. This CBC analysis suggests that parents of children with EOS requiring surgical intervention would prioritize complication minimization, reduction in total surgical procedures, and effective curve correction, demonstrating a marked preference for PSF. Notably, the sex of both the child and the parent influenced attribute preferences. These findings exemplify the utility of CBC analysis in orthopaedic care, allowing for a deeper understanding of family preferences and enhancing SDM by providing the ability to align surgical recommendations with specific family priorities. Level III.

Discrete choice modeling (DCM) is widely used in economics, social studies, and marketing research for estimating utilities and preference probabilities of multiple alternatives. Data for the model is elicited from the respondents who are presented with several sets of items characterized by various attributes, and each respondent chooses the best alternative in each set. Estimation of utilities is usually performed in a multinomial-logit (MNL) modeling and software for Hierarchical Bayesian (HB) technique is usually applied to find individual utilities by iterative estimations. This paper describes an easy and convenient empirical Bayesian way to construct priors and combine them with the likelihood on individual level data. This allows the modeler to obtain posterior estimation of MNL utilities in noniterative evaluations. Logistic modeling for the posterior frequencies is performed using the linear link of their logarithm of odds that clarifies the results of DCM modeling. The problem of overfitting is considered and the optimum balance between signal and noise in the precision of individual prediction and the smoothing of overall data is suggested. Actual market research data are used and the results are discussed.

This paper presents estimations of willingness-to-pay (WTP) values in the context of rural bus service in a developing country. Using stated choice data collected from rural bus users, WTP values are estimated by developing standard multinomial logit model (MNL) and three different random parameter logit (RPL) models: One with independent or uncorrelated responses, another taking into account the correlations among responses, and the other allowing heterogeneity around the mean of in-vehicle travel time. While developing the RPL models, successful application of the sparsely used constrained triangular distribution is demonstrated. WTP values are estimated separately for commuting and noncommuting trips. Higher WTP values for travel time are observed for commuting trips, while for noncommuting trips higher WTP values are found for qualitative attribute. Preference heterogeneity associated with the mean is investigated, and the “travel distance” is found to have a statistically significant decomposition effect on the mean of in-vehicle travel time for commuting trips. The traditional MNL model, in general, is found to underestimate the WTP values as compared to RPL models.

In response to consumers’ increasing concerns about food safety, both supply chain operators and public authorities have developed, in the last decade, labelling schemes to provide markets with more information. Although some level of traceability is necessary to provide consumers with valuable information (such as country-of-origin, production process, ecc.), it is not clear so far if traceability per se is important to consumers. The objective of this paper is to analyse consumers’ perception of food traceability with respect to two food products: chicken and honey. More specifically, it aims to provide more insights for both policy makers and supply chain actors into Italian consumer’s perception and preferences for different attributes connected to traceability. After a preliminary qualitative phase, a choice experiment (ce) was employed on a sample of Italian consumers to analyse preferences for different profiles of chicken and honey. ce methodology allows to weight the relative importance of any given attribute for consumers, providing an estimation of their marginal willingness to pay (wtp). A multinomial logit (ml) and a latent class (lc) model were applied to estimate the parameters associated to each product’s attribute. The lc model, in particular, fitted better the data providing more detailed results able to capture heterogeneity of preferences across the considered products. One of the findings in this study is that traceability perception is a product specific issue: consumers have more preferences for traceability attributes in the case of chicken than for honey. This is probably due to a different level of perceived risk and to the higher market penetration of chicken. Therefore, on the one hand, we suggest supply chain operators to develop traceability-based valorisation strategies per product. On the other hand, the introduction of vertical public regulations on food traceability, i.e. defined per product category, should be preferred by public authority.