Latent Class Regression Research Articles

The role of attitudes in perceptions of bicycle facilities: A latent-class regression approach

While many professionals support bicycling for transportation to reduce traffic congestion and improve public health, many potential cyclists see the risks from traffic injuries as a major barrier. The purpose of this paper is to investigate the impact of bicycle facilities on perceptions of bikeability, and how these perceptions vary among different groups. Data comes from surveys (N = 2157) deployed in ten study neighborhoods as part of two research projects. Respondents were asked to rate a series of images of hypothetical roadways in terms of perceived comfort, perceived safety, and willingness to try bicycling. Latent-class regression was used to model these responses, with bicycle facility type and roadway characteristics as explanatory variables and sociodemographics and attitudinal factors (car preference, bike enjoyment, risk tolerance, and anti-exercise) as covariates for class membership. A recurring theme among the presented solutions is the emergence of two pro-bicycle classes - risk-embracing and risk-cautious - together with a pro-car class. Results indicate that the impact of protected bicycle facilities on perceptions are much higher for the pro-bike/risk-cautious group. The major implication of these findings is that there likely exists a sizeable segment of the population that, although perhaps just as interested in bicycling as the typical pro-bicycle class, is more cautious in their perceptions of comfort and safety. Those belonging to this subgroup generally view minimal facilities as much more unsafe and uncomfortable than their risk-accepting counterparts, but have comparable perceptions regarding high-quality protected bicycle facilities.

Penerapan Latent Class Regression Analysis dalam Segmentasi Pasar

Cluster analysis is a method of grouping observation objects into several classes. One method of mixed-scale data grouping is Latent Class Regression Analysis (LCRA). The purpose of this research is to classify the opinion of Wardah's product consumers on marketing strategies (product aspects, price aspects, location aspects, and promotion aspects) PT. Paragon Technology and Innovation Regional Samarinda in 2017 with covariate variables arelength of subscription, type of work and age of consumers. Estimation of LCRA using the Expectation Maximization (EM) method, solved by the Newton-Raphson method. The result of LCRA analysis that based on consumer opinion on market segmentation, consumers are grouped into two classes.The first class is 31 consumers that strongly agrees the aspects of product, price, promotion and position are appropriate market segmentation, and the second class is 69 that quite agrees product aspects, prices, promotion and position is the appropriate market segmentation.

Clustering of multiple health risk factors among a sample of adolescents in Liberia: a latent class analysis

Non-communicable diseases (NCDs) are associated with modifiable health risk factors. There is a lack of evidence regarding clusters of health-related behaviours among school-going adolescents from sub-Saharan Africa. This study was conducted to identify clustering patterns of health risk factors (i.e. smoking tobacco, inadequate fruit intake, inadequate vegetable intake, physical inactivity, sedentary behaviour, anxiety and depression) and association with sociodemographic factors among school-going adolescents in Liberia. The 2017 Liberian Global School-based Student Health Survey on 2774 adolescents aged 11 years and above (52.5% females) sampled with a two-stage cluster sample design was used. Latent class analysis was used to generate the clusters and latent class regression assessed the associations between sociodemographic factors and the clusters. We identified three clusters labelled as (1) ‘low substance use, moderately active cluster’ (34.8%); (2) ‘inadequate fruit and vegetable cluster’ (48.9%) and (3) ‘risk taking cluster’ (16.3%)’. Compared to cluster 1, adolescent boys [AOR = 1.71, 1.29–2.27, p < 0.001], and those in grade 10–12 [AOR = 1.51, 1.13–2.02, p < 0.001] were more likely to belong to cluster 2. Participants aged 15 years and above [AOR = 0.60, 0.39–0.91, p = 0.018] were less likely to belong to cluster 2. Compared to cluster 1, adolescents aged 15 years and above [AOR = 3.58, 1.33–9.62, p = 0.011] and those with low socio-economic status [AOR = 1.83, 1.22–2.73, p = 0.003] were more likely to belong to cluster 3. These results underscore the need for interventions that address adolescent multiple health risk factors, especially considering sociodemographic differences.

Personalized Ranking of Online Reviews Based on Consumer Preferences in Product Features

ABSTRACT Online consumer reviews (OCRs) can function as a venue for digital collaboration among various stakeholders to better meet collaborate in consumer needs. The sheer volume of OCRs, however, has posed challenges to efficient search and navigation. Importantly, consumers' needs of product information may differ because of their different preferences in product features. Such differences remain underaddressed in the OCR literature. This research introduces a novel framework - Product feature based Personalized Review Ranking (P2R2), which predicts review helpfulness for individual consumers based on their preferences in product features using a latent class regression model. The framework also leverages the similarities among different consumers to derive consumer classes. An empirical evaluation of a prototype of P2R2 with a user study provides strong evidence that the review rankings produced by P2R2 are more similar to users’ self-rankings than by a helpfulness vote based ranking method. The findings of this study offer theoretical insights, novel technical design artifacts, and empirical evidence for enhancing OCR platforms with review ranking personalization.

Analysis of subway users' behaviour-based on the latent class regression

Analysis of subway users' behaviour-based on the latent class regression

Analysis of subway users' behaviour-based on the latent class regression

Analysis of subway users' behaviour-based on the latent class regression

Empathy profiles differ by gender in people who have and have not attempted suicide

High levels of empathy may inhibit hetero-aggressive behaviors; however, the role of empathy on suicide behaviors is still unknown. This study aimed to compare the Interpersonal Reactivity Index (IRI) empathic profiles of 56 patients hospitalized after a suicide attempt with those of 138 people who had never attempted suicide. Differences were tested with t-test and Chi-square test. The associations between attempted suicide and empathy scores were tested with linear regression models, controlling for sex and age. Latent Class Regression Analysis was applied to investigate the relationship between multivariate categorical empathy response items and suicide attempts, controlling for sex and age. Suicide attempters scored significantly higher on the ‘Personal Distress’ and ‘Fantasy’ IRI subscales. Women in the control group had similar probabilities of belonging to the class of high or low Personal Distress and Fantasy levels, while women who had attempted suicide were more likely to have high scores at the same scales. Men in the control group had higher probabilities of scoring low at the Personal Distress and Fantasy subscales, while men who had attempted suicide had similar probabilities of belonging to the class with high or low scores. The use of a self-administered tool may have introduced a gender-role stereotype bias in empathy assessment. Additionally, it was not possible to test the role of psychopathology. Our results suggest that distinct empathic profiles are associated with suicidal behavior, with a gender specific pattern. Addressing empathy constructs may help identifying suicidal individuals.

Analysing Consumer Preferences, Characteristics, and Behaviour to Identify Energy-Efficient Consumers

This paper investigates preference heterogeneity among Slovenian energy consumers and attempts to ascertain how different consumer groups value various attributes of energy products and services. More specifically, it aims to establish whether a consumer segment can be identified that shows a preference for additional energy services—in particular services, associated with energy-efficient and green behaviour. A latent class analysis is employed to classify consumers on the basis of their preferences for energy services. Additionally, information about their attitudes and behaviour toward green energy and energy efficiency, energy consumption, and usage of energy services together with socio-economic characteristics is used in the latent class regression to explain differences between latent consumer classes. Three classes are identified: the largest class of regular consumers, energy-efficient consumers, and dissatisfied consumers. In contrast to regular and dissatisfied consumers, energy-efficient consumers show a significantly higher interest in additional services, energy efficiency, and green energy. In line with the found heterogeneity of consumer preferences, suppliers should customise marketing strategies to meet the needs of specific segments. Energy policymakers also need to pay more attention to consumer heterogeneity and behavioural changes to increase the effectiveness of energy efficiency policies.

Latent Class Regression Utilizing Fuzzy Clusterwise Generalized Structured Component Analysis

Latent class analysis (LCA) has been applied in many research areas to disentangle the heterogeneity of a population. Despite its popularity, its estimation method is limited to maximum likelihood estimation (MLE), which requires large samples to satisfy both the multivariate normality assumption and local independence assumption. Although many suggestions regarding adequate sample sizes were proposed, researchers continue to apply LCA with relatively smaller samples. When covariates are involved, the estimation issue is encountered more. In this study, we suggest a different estimating approach for LCA with covariates, also known as latent class regression (LCR), using a fuzzy clustering method and generalized structured component analysis (GSCA). This new approach is free from the distributional assumption and stable in estimating parameters. Parallel to the three-step approach used in the MLE-based LCA, we extend an algorithm of fuzzy clusterwise GSCA into LCR. This proposed algorithm has been demonstrated with an empirical data with both categorical and continuous covariates. Because the proposed algorithm can be used for a relatively small sample in LCR without requiring a multivariate normality assumption, the new algorithm is more applicable to social, behavioral, and health sciences.

Longitudinal Associations Between Sexual Victimization and Substance Misuse Among High School Youth in Colorado.

There are consistently high rates of sexual victimization and substance misuse among youth in the United States. Although there is a known relation between sexual victimization and substance use, there is a gap in the research regarding the magnitude and temporality of these associations. This study examined whether latent classes of sexual victimization and help-seeking attitudes longitudinally predict intentions of future substance misuse 7-10 months later. Students from nine Colorado high schools (N = 533; M = age 16 years) completed surveys across two consecutive school semesters. Latent class analysis was used to identify classes of students who experienced at least one form of sexual victimization (e.g., sexual harassment, unwanted sexual contact) according to 13 sexual victimization items, and level of positive attitudes regarding help-seeking for sexual victimization. Classes were compared on demographic characteristics and for distal outcomes of likelihood of future substance misuse (cigarettes, alcohol, cannabis, prescription drugs, and electronic vaping products) using latent class regression, controlling for previous intentions to use. At Time 1, four classes of sexual victimization were identified with two main classes for comparison: low odds of experiencing sexual victimization (60.1% of sample) and high odds of endorsing all forms of sexual victimization (7.7% of sample). The high sexual victimization class had higher proportions of male and transgender students, compared to other classes. At Time 2 (7-10 months later), students in the high sexual victimization class reported a significantly greater likelihood of future cigarette (p = .017) and prescription drug misuse (p = .007) when compared to the low sexual victimization class. There was no evidence that having higher positive attitudes towards help-seeking resulted in lower intentions to use substances in the future. These findings highlight that addressing sexual violence in prevention programming may have crossover effects of reducing substance misuse and other forms of violence among youth.

Patterns of physical activity, sitting time, and sleep in Australian adults: A latent class analysis

ObjectiveTo identify the patterns of activity, sitting and sleep that adults engage in, the demographic and biological correlates of activity-sleep patterns and the relationship between identified patterns and self-rated health. Design and SettingOnline panel of randomly selected Australian adults (n = 2034) completing a cross-sectional survey in October-November 2013. ParticipantsPanel members who provided complete data on all variables were included (n = 1532). MeasurementsParticipants self-reported their demographic characteristics, height, weight, self-rated health, duration of physical activity, frequency of resistance training, sitting time, sleep duration, sleep quality, and variability in bed and wake times. Activity-sleep patterns were determined using latent class analysis. Latent class regression was used to examine the relationships between identified patterns, demographic and biological characteristics, and self-rated health. ResultsA 4-class model fit the data best, characterized by very active good sleepers, inactive good sleepers, inactive poor sleepers, moderately active good sleepers, representing 38.2%, 22.2%, 21.2%, and 18.4% of the sample, respectively. Relative to the very active good sleepers, the inactive poor sleepers, and inactive good sleepers were more likely to report being female, lower education, higher body mass index, and lower self-rated health, the moderately active good sleepers were more likely to be older, report lower education, higher body mass index and lower self-rated health. Associations between activity-sleep pattern and self-rated health were the largest in the inactive poor sleepers. ConclusionsThe 4 activity-sleep patterns identified had distinct behavioral profiles, sociodemographic correlates, and relationships with self-rated health. Many adults could benefit from behavioral interventions targeting improvements in physical activity and sleep.

Predicting problematic smartphone use over time in adolescence: A latent class regression analysis of online and offline activities

Despite today’s ubiquitous nature of smartphones among adolescents, little is known about behavioural online and offline longitudinal predictors of problematic smartphone use (PSU). Guided by Uses and Gratifications Theory, we applied latent class analysis on survey data collected in 2017 from a cohort of 1096 adolescents ( Mage = 12.4, SDage = 0.56) and regressed PSU measured 1 year later on class membership, controlling for socio-demographic characteristics, social desirability and autoregressive effects. We extracted four distinct classes: social-recreational onliners ( n = 228), weekend onliners ( n = 331), balanced ( n = 404) and noninvolved ( n = 153). Characterised by significantly more time spent online for recreational and social networking activities, both during weekdays and weekend days, as well as less time for sleep, the social-recreational onliners class showed significantly higher levels of PSU over time. Future studies should assess not only duration but also the frequency of daily online activities to provide further insights into behavioural predictors of PSU.

How Different Stressors Affect Quality of Life: An Application of Multilevel Latent Class Analysis on a Large Sample of Industrial Employees.

BackgroundQuality of life (QoL) indicates individual’s perception of the physical, psychological and social aspects of health. The association between QoL and all stressful life events‘ dimensions has not been investigated among industrial employees.PurposeThe present study aimed to investigate the association between stressful life events and QoL in a large sample of Iranian industrial employees.Material and MethodsIn a cross-sectional study, 3063 manufacturing employees in Isfahan, Iran, were recruited. QoL was measured with the Euro-QoL five dimension questionnaire (EQ-5D). Stressful life events were assessed by an Iranian validated stressful life event questionnaire (SLE). Multilevel latent class regression was employed for classifying participants based on QoL and for determining its association with stressful life events.ResultsTwo classes of employees, ie, low (comprised 20% of participants) and high QoL (80%), were identified. From 11 dimensions of stressful life events two major domains, socioeconomic and personal stressor profiles, were identified. Multilevel latent class regression results showed that the higher scores of social (OR: 1.14, 95%CI: 1.12–1.16) and personal (OR: 2.36, 95%CI: 1.87– 2.98) stressor domains were significantly associated with increased risk of being in the poor QoL class. Among personal and socioeconomic stressors, health concerns and daily life dimensions had higher significant association with poor QoL.ConclusionResults of our study indicated that life stressors have negative impacts on QoL of employees. Personal stressors showed higher negative impact than socioeconomic on QoL. Managing the stressors can play an effective role in improving the QoL of employees, their physical and mental health, and indirectly enhances the organizational performance and job productivity.

Hands-On Exploration of Cubes' Floating and Sinking Benefits Children's Subsequent Buoyancy Predictions.

Children accrue experiences with buoyancy on a daily basis, yet research paints a mixed picture of children’s buoyancy knowledge. Whereas children’s predictions and explanations of the floating and the sinking of common objects are often based on a single feature (e.g., mass or facts), children’s predictions of novel cubes reveal solution strategies based on mass and volume integrations. Correspondingly, category learning theory suggests that categories (e.g., floaters vs. sinkers) are easier to identify when items mainly vary from one another in the relevant defining features. For example, a set of cubes only varies in mass and volume and hence density, thereby being able to highlight the deterministic role of density when placed in water. Here we asked how item variation during hands-on exploration affects children’s subsequent predictions and explanations of buoyancy. Kindergarteners and first-, second-, and third-grade children individually explored either a set of 10 systematically varied cubes (i.e., systematic condition; n = 95) or a set of 10 common objects (i.e., non-systematic condition; n = 96) in a water basin. Next, the children predicted the buoyancy of five new cubes and five new common objects one at a time. Subsequently, the children explained their predictions one subset at a time. The children in the systematic condition were more accurate in their predictions of the test cubes than the children in the non-systematic condition. Latent class regression analyses identified three cube prediction solution strategies. The children in the systematic condition were more likely to use a strategy in which buoyancy decisions were made based on an accurate integration of mass and volume, while the children in the non-systematic condition were more likely to use a strategy in which mass was given more predictive load than volume. A third strategy was characterized by guessing. Latent class analyses of the children’s explanations revealed different explanation strategies, each appealing to several features, but as hypothesized, no clear condition differences were found. The findings indicate that even 5 min of exploration with systematically varied cubes can already help children use an advanced buoyancy prediction strategy. This provides evidence in favor of using category learning theory to inform early science education design.

Gender and Sexual Identities Predicting Patterns of Co-occurring Health Risks Among Sexual Minority Youth: a Latent Class Analysis Approach.

Behavioral health disparities (e.g., substance use, mental health) exist for gender nonconforming (GNC) and sexual minority youth; however, there is limited knowledge on disparities that may be unique among youth who identify as both a sexual and gender minority. This study utilized a diverse sample of GNC and cisgender sexual minority youth seeking crisis services to examine co-occurrence of behavioral health outcomes. Surveys were administered with youth (aged 12-24, N = 592), and latent class analyses were applied. Two latent class regression models were conducted to examine how gender and sexual identity separately (independent effect; Model 1) and configurations of gender and sexual identity (Model 2) predicted class membership. Analyses resulted in a four-class solution: High All (17.6%), High Substance Use and Moderate Mental Health (10.6%), Low All (20.1%), and High Suicide and High Mental Health (51.7%). In our first model, youth who identified as GNC had 2.11 higher odds of being in the High Suicide and High Mental Health class compared to the Low All class; however, sexual identity was not a significant predictor. In the second model, individuals identifying as GNC gay or lesbian or GNC pansexual had 1.95 and 2.57 higher odds, respectively, of being in the High Suicide and High Mental Health class compared to the Low All class. Our study suggests the information on gender and sexual identities together are more helpful in identifying youth at risk for co-occurring negative health outcomes. Implications for prevention approaches are described.

Substance Use Profiles Among Urban Adolescents: The Role of Family-Based Adversities

The purpose of this study is to advance our understanding of family-based adversities and adolescent well-being by examining variability in substance use profiles among urban, low-income and predominantly African American adolescents (n = 2858). Latent class analysis and regressions were used to identify the nature and correlates of heterogeneity in substance use patterns among adolescents in the Mobile Youth Survey (MYS). Analyses revealed the presence of four underlying subgroups of youth: non-users (48.4%), alcohol users (17.6%), alcohol and marijuana users (31.6%), and polysubstance users (2.5%). Membership in classes varied by age and gender, with older youth and males being more likely to belong to groups characterized by the use of multiple substances. Youth who reported that someone in their household had been arrested were more likely to belong to one of the substance using groups than the non-use group, and youth who reported not having a father figure were more likely to belong to groups characterized by the use of multiple substances relative to the non-use group. Youth who reported living with two biological parents were less likely to belong to either the alcohol or the alcohol and marijuana classes than the non-use group. Results further indicated that youth in classes characterized by the use of multiple substances were more likely to exhibit externalizing problems and internalizing distress than other youth. Findings underscore the importance of identifying heterogeneity even in seemingly homogeneous samples and suggest a number of new directions for future research.

Accumulated unhealthy behaviors and psychosocial problems in adolescence are associated with labor market exclusion in early adulthood \u2013 a northern Finland birth cohort 1986 study

BackgroundThe relevance of health-related behaviors to exclusion from the labor market in early adulthood remains poorly studied in relation to the magnitude of the problem. We explored whether adolescents’ accumulated unhealthy behaviors and psychosocial problems are associated with later labor market exclusion, and whether multisite musculoskeletal pain (MMSP) impacts these relations.MethodsWe gathered questionnaire data on unhealthy behaviors and psychosocial problems and MMSP among adolescents aged 15 to 16 belonging to the Northern Finland Birth Cohort 1986. The findings were combined with registry data on unemployment, employment and permanent work disability during a five-year follow-up between the ages of 25 and 29 (n = 6692). In the statistical modeling we used education, family leave and socioeconomic status of childhood family as potential confounders, as well as latent class and logistic regression analyses.ResultsThe Externalizing behavior cluster associated with over one year of unemployment (RR 1.64, CI 1.25–2.14) and permanent work disability (OR 2.49, CI 1.07–5.78) in the follow-up among the men. The Sedentary cluster also associated with over one year (RR 1.41, CI 1.13–1.75) and under one year of unemployment (RR 1.25, CI 1.02–1.52) and no employment days (RR 1.93, CI 1.26–2.95) among the men. Obese male participants were at risk of over one year of unemployment (RR 1.50, CI 1.08–2.09) and no employment days (RR 1.93, CI 1.07–3.50). Among the women, the Multiple risk behavior cluster related significantly to over one year of unemployment (RR 1.77, CI 1.37–2.28). MMSP had no influence on the associations.ConclusionsUnhealthy behavior patterns and psychosocial problems in adolescence have long-term consequences for exclusion from the labor market in early adulthood, especially among men. Simultaneously supporting psychological well-being and healthy behaviors in adolescence may reduce labor market inclusion difficulties in the early phase of working life.

Self-rated health in adolescence as a predictor of 'multi-illness' in early adulthood: A prospective registry-based Norwegian HUNT study.

Self-rated health (SRH) is a marker of future health and a possible predictor of future multimorbidity, which is a major challenge for population health and health care. There is a lack of studies on adolescent SRH and patterns of health problems across the transitional period from adolescence to early adulthood. Therefore, this study aimed to identify groups of people with similar health problems in early adulthood and explore the predictive value of adolescent SRH on the group classification after a period of 10–19 years. Data from 8828 adolescents participating in the Young HUNT-1 survey (1995–1997) were linked to the Norwegian registry of general practitioner (GP) claims, which includes diagnoses recorded in GP consultations in 2006–2014. We used latent class analysis (LCA) to identify groups of patients with similar health problems in early adulthood and explored SRH as a predictor of class membership using latent class regression, adjusting for baseline chronic disease, frequency of health care attendance, sex and age. The mean age at baseline was 16 years, and 50% of the participants were female. SRH was reported as very good by 28%, good by 61% and not good by 11%. We identified five groups of patient classification (classes): Healthy (35%), Infections and general problems (26%), Musculoskeletal problems (21%), Psychological problems (6%) and Multi-illness (13%). We found a gradual increase in the probability of belonging to the Healthy class with better SRH, and an inverse pattern for the Psychological and Multi-illness classes. This pattern remained after adjusting for baseline variables. In conclusion, there is a clear association between adolescent SRH and the risk of having multi-illness in early adulthood, seen as a proxy for later multimorbidity. This finding warrants greater attention to SRH in adolescence as a possible indicator in targeted prevention of future health problems.

Patterns of Exposure to Socio-structural Stressors and HIV Care Engagement Among Transgender Women of Color.

Transgender women are disproportionately affected by HIV and experiences of social adversity that may interfere with engagement in care and viral suppression. We used latent class analysis to examine patterns of social adversity and their impact on HIV care continuum outcomes in an urban sample of transgender women of color. Participants (n = 224) were median age 29 and 86% non-Hispanic Black. Lack of resources, unemployment, and housing instability were reported by over 50%, and 41% reported history of incarceration. Latent class analysis identified 2 distinct classes representing higher and lower levels of social adversity. In latent class regression, membership in the higher social adversity class was associated with statistically significantly lower odds of viral suppression and HIV care engagement in univariate analysis; when adjusted for age, race, and recruitment site the association remained statistically significant for viral suppression (aOR 0.38, 95% CI 0.18-0.79; chi-square = 6.681, d.f. = 1, p = 0.010), though not for HIV care engagement. Our findings highlight the impact of socio-structural barriers on engagement in the HIV care continuum among transgender women.

The role of women on boards in corporate environmental strategy and financial performance: A global outlook

AbstractThis study examines the impact of board gender diversity on corporate environmental strategy and financial performance. Based on 12 corporate environmental policies in 3,389 firms worldwide, we identified four types of corporate environmental strategies by using the latent class regression model: an inactive strategy, a reactive strategy, a pollution prevention strategy, and a sustainable development strategy. The empirical evidence shows that women on boards (WoB) contribute to the promotion of proactive environmental strategies, including the pollution prevention strategy, which is found to bring about sustained competitive advantage in both short‐term and long‐term financial performance, and the sustainable development strategy, which is positively associated with long‐term financial performance. Following the natural resource‐based view of the firm, these findings indicate that WoB can be seen as a key resource in the organizational process, which provides a shared vision of the future and strong moral leadership to the top management team.