Random Effects Negative Binomial Model Research Articles

Mosquito Bed Net Use and Burkitt Lymphoma Incidence in Sub-Saharan Africa

Burkitt lymphoma (BL) is one of the most common childhood cancers in sub-Saharan Africa and is etiologically linked to malaria. However, evidence for an effect of malaria interventions on BL is limited. To investigate the potential population-level association between large-scale rollout of insecticide-treated bed nets (ITNs) in sub-Saharan Africa in the 2000s and BL incidence. In this systematic review and meta-analysis, a search was conducted in the Embase, Global Health, and Medline databases and in cancer registry publications between January 1, 1990, and February 27, 2023. All epidemiologic studies on BL incidence rates in children and adolescents aged 0 to 15 years in sub-Saharan African countries where malaria is endemic were identified by 2 reviewers blinded to each other's decision. The systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses reporting guideline. Data were extracted independently by 2 reviewers, and quality was scored based on 3 predefined criteria: data collection, case ascertainment, and calculation of person-time at risk. Incidence rates of BL during childhood and mean ITN use in the population. Data were analyzed using a random-effects negative binomial regression model. Of 2333 studies meeting selection criteria, 23 comprising 66 data points on BL incidence were included based on 5226 BL cases from locations with large-scale ITN use in 17 countries. Rates of BL were 44% (95% CI, 12%-64%) lower in the period after ITN introduction compared with before. The adjusted pooled incidence rates of BL were 1.36 (95% CI, 0.88-2.10) and 0.76 (95% CI, 0.50-1.16) per 100 000 person-years before and after introduction of ITNs, respectively. After adjusting for potential confounders, a 1-percentage point increase in mean ITN use in the population in the 10 years before BL data collection was associated with a 2% (95% CI, 1%-4%) reduction in BL incidence. In this systematic review and meta-analysis, large-scale rollout of ITNs in the 2000s was associated with a reduction in BL burden among children in sub-Saharan Africa. Although published data may not be representative of all incidence rates across sub-Saharan Africa, this study highlights a potential additional benefit of malaria control programs.

Freeway crash risk prediction considering unobserved heterogeneity: A random effect negative binomial regression approach

Unobserved heterogeneity of crashes remains a significant issue for freeways that influence crash prediction, and therefore deserves much attention. Using a fusion data set of crash data, driving behavior data, and traffic flow data, this study explores the spatiotemporal heterogeneity of crash determinants for different freeway segments (e.g. Yixing section and Liyang section of Ning-Hang freeway of China) and then predict the crash probability. A random effect negative binomial regression model is built to investigate the contributing factors of the crashes. Remarkable differences are observed in the crash determinants for Yixing section (include average vehicle speed, hourly average traffic volume, average free speed, road segment length, and number of left lane-merging) and Liyang section (include average intensity of aggressive driving behavior, average kilometer traffic volume). The results found the traffic flow has a more significant impact on crashes than the driving behaviors. It is found that the crash probability is a monotone decreasing function when the predicted number of crash is 0. With the increase of the number of predicted crash, the crash probability gradually converges from a large value to 0. Then the probability of other predicted number of crashes (e.g. crash = 1, crash = 2, crash = 3) presents a quadratic parabola trends. The model comparison demonstrates that the proposed model outperforms conventional model, and the prediction performance for Liyang section is better than that of Yixing section. The research findings are interesting and important for preventing crashes.

Assessing the Effect of COVID-19 on the Traffic Safety of Intercity and Major Intracity Roads in Saudi Arabia.

The coronavirus disease 2019 (COVID-19) pandemic led to rapid and unexpected changes across the world, particularly in road safety. Thus, this work assesses the impact of COVID-19 accompanied by government preventive policies on road safety in Saudi Arabia by investigating the crash frequency and crash rates. A 4-year crash dataset relating to 2018-2021 was collected, covering about 71,000km in total road length. It covers all intercity roads and some of the major intercity roads in Saudi Arabia with over 40,000 data logs of involved crashes. We considered three different time phases to observe road safety. These time phases were identified by the duration of government curfew measures against COVID-19 (before, during, and after). The crash frequency analysis showed that the curfew during COVID-19 significantly impacted the crash reduction. At a national level, the crash frequency decreased during 2020 and reached a 33.2% reduction compared to 2019 (the previous year), and it surprisingly continued decreasing in 2021 (the consequent year) to another 37.7% reduction although the government measures were lifted. Moreover, considering the traffic volume and road geometry, we analyzed crash rates for 36 selected segments, and the results showed a significant reduction in the crash rate before and after the COVID-19 pandemic. Additionally, a random effect negative binomial model was developed to quantify the impact of the COVID-19 pandemic. The results showed that the reduction in crashes was significant during and after COVID-19. Also, single roads (two-lane, two-way) were found to be more dangerous than other types of roads.

Investigating factors affecting rural crash frequency at marginal areas around cities in Iran: a random parameter modeling approach

ABSTRACT A significant portion of rural crashes in Iran occur in the marginal areas around cities. Exclusive crash frequency models should be developed to identify the factors affecting the occurrence of crashes in these areas. In this study, the rural crash data of highways leading to the city of Isfahan were employed to develop a negative binomial (NB) model as a constant-parameter model, along with random effects negative binomial (RENB) and random parameters negative binomial (RPNB) models as the models based on the random parameter approach. The modeling results indicated the superiority of the RPNB model over the two other models. According to the findings of this study, the traffic volume of roads and the proportion of local and non-local vehicles, the proportion of longitudinal distance-limit violation, the way of controlling access points, and the access to rest areas affected the crash frequency in the marginal areas around cities.

How is traffic safety affected by changes in traffic speeds following speed limit increases? An evaluation with probe vehicle data

Maximum speed limits continue to be an important policy issue. Research has consistently shown speeds to increase following speed limit increases. These increases have also generally coincided with increases in both the frequency and severity of crashes. However, research has been more limited as to the direct relationship between speed and safety. To this end, the present study evaluates the effects of increasing speed limits on the safety of rural limited access freeways while accounting for contemporaneous changes in the speed profiles of these same roadways. The speed limit increases in Michigan occurred between May 1 and June 12 of 2017, when the maximum speed limits were increased from 70 to 75 mph on approximately 600 miles of freeways. The maximum speed limits for trucks were also increased from 60 to 65 mph on all freeways state-wide at this same time. Speed data were obtained for the Michigan rural freeway network through probe vehicles. These data are merged with pertinent roadway data, as well as police-reported crash data at various levels of injury severity. The impacts of the speed limit increase on safety were evaluated by estimating random effects negative binomial models as part of a case-control before-after study design. The results show that the locations where the speed limits were increased experienced a 5% increase in crashes, while a marginal reduction in crashes was observed where speed limits did not increase. Interestingly, mean speeds were found to be negatively associated with crash frequency, while the standard deviation of speed was found to exhibit a positive relationship. Several site-specific characteristics were also found to be strong predictors of crash frequency. The results provide important insights into the nature of the relationship between speed and safety and will help to guide subsequent speed limit policy decisions.

Equity in the use of physician services in Canada's universal health system: A longitudinal analysis of older adults

This study presents longitudinal evidence on the trends and determinants of income-related inequities in general practitioner (GP), specialist, and any physician visits among older adults (aged 65+) in Canada. Using the Canadian National Population Health Survey between 1998/99 and 2010/11, random effect probit and negative binomial models were employed to model the probability of visit and the total number of visits, respectively. The concentration index-based horizontal inequity (HI) approach was used to measure income-related inequities in physician services. The decomposition technique was applied to explain the factors contributing to the observed inequities. The mobility index (MI) was also calculated to compare short-run and long-run estimates of inequities. The HI indices reveal significant pro-rich inequities in both the probability and the number of specialist visits. Inequities in the likelihood of GP visits and any physician visits were pro-rich but trivial in magnitude. The MI shows that upwardly income mobile individuals contribute to inequity in specialist visits in the long run. After income, education was the most important contributor to inequity in specialist visits, while unobserved heterogeneity explained most of the pro-rich inequity in the total number of GP and any physician visits. Although physician services are free at the point of the provision in Canada, this study demonstrates that poorer older adults utilized fewer specialist services than richer older adults for the same level of need. Specific policies are needed to ensure equity in specialist care use among the older adults in Canada.

Accident Frequency Prediction Model for Flat Rural Roads in Serbia

Traffic accidents, by their nature, are random events; therefore, it is difficult to estimate the exact places and times of their occurrences and the true nature of their impacts. Although they are hard to precisely predict, preventative actions can be taken and their numbers (in a certain period) can be approximately predicted. In this study, we investigated the relationship between accident frequency and factors that affect accident frequency; we used accident data for events that occurred on a flat rural state road in Serbia. The analysis was conducted using five statistical models, i.e., Poisson, negative binomial, random effect negative binomial, zero-inflated Poisson, and zero-inflated negative binomial models. The results indicated that the random effect negative binomial model outperformed the other models in terms of goodness-of-fit measures; it was chosen as the accident prediction model for flat rural roads. Four explanatory variables—annual average daily traffic, segment length, number of horizontal curves, and access road density—were found to significantly affect accident frequency. The results of this research can help road authorities make decisions about interventions and investments in road networks, designing new roads, and reconstructing existing roads.

Safety Evaluation of Freeway Exit Ramps with Advisory Speed Reductions

Posted speed limits inform drivers of the legal maximum allowable speed. In contrast, advisory speeds provide recommendations to drivers based on highway design, operating characteristics, and conditions. Various studies have investigated the safety impacts of speed limit changes, particularly on high-speed rural highways. One area of particular concern on such roadways is the approach to exit ramps that require substantive speed reductions, such as loop ramps. To date, there has been limited research examining the safety impact of the differential between the mainline speed limit and the lower exit ramp advisory speeds. This study aims to evaluate this relationship through the estimate of a series of safety performance functions. Random effects negative binomial regression models were estimated using data from 187 exit ramps where advisory speed signs are present throughout Michigan. The analyses were conducted based on a five-year period from 2014 to 2019, excluding 2017 data because of speed limit increases. In addition to speed differentials, various roadway characteristics such as deceleration lane length and curve radius were also evaluated. Results indicate that crashes increased as the difference between mainline and ramp speed increased. Crashes were also shown to decrease on ramps where the upstream deceleration lane length was greater than the minimum recommendation as per current design guidance.

The Impact of COVID-19 on Gun Violence Across Census Tracts in NYC

This study explores the impact of COVID-19 on gun violence in NYC and its interactive effects with neighborhood factors at the census tract level. Random effects negative binomial models are used to analyze monthly data from January 2017 to March 2021. There was a significant increase in gun violence during the pandemic. In addition, poverty, economic inequality, African Americans, Hispanics, residential mobility, and total population were significantly associated with increases in gun violence. However, there were no significant interaction effects between the pandemic and neighborhood characteristics. This study concludes with a discussion of study limitations and implications.

Impacts of Real-Time Traffic State on Urban Expressway Crashes by Collision and Vehicle Type

With the rapid development of urban expressway systems in China in recent years, traffic safety problems have attracted more attention. Variation of traffic flow is considered to have significant impact on the safety performance of expressways. Therefore, the motivation of this study is to explore the mechanism of how the variation of traffic flow measurements such as average speed, speed variation and traffic volume impact the crash risk. Firstly, the crashes were classified according to crash type and vehicles involved: and they are labeled with rear-end collisions or side-impact collisions, they are labeled with heavy-vehicle related collisions or light-vehicle related collisions as well. Then, the corresponding crash data were aggregated based on the similarity of traffic flow conditions and types of crashes. Finally, a random effect negative binomial model was introduced to consider the heterogeneity of the crash risk due to the variance within the traffic flow and crash types. The results show that the significant influencing factors of each type of crashes are not consistent. Specifically, the percentage of heavy vehicles within traffic flow is found to have a negative impact on rear-end collisions and light-vehicle-related collisions, but it has no obvious correlation with side-impact collisions and heavy-vehicle-related collisions. Average speed, speed variation and traffic volume have an interactive effect on the crash rate. In conclusion, if the traffic flow is with higher speed variation within lanes and is with lower average speed, the risk of all types of crashes tends to be higher. If the speed variation within lanes decreases and the average speed increases, the crash risk will also increase. In addition, if the traffic flow is under the conditions of higher speed variation between lanes and lower traffic volume, the risk of rear-end collisions, side-impact collisions and heavy-vehicles related collisions tend to be higher. Meanwhile, if the speed variation between lanes decreases and the traffic volume increases, the crash risk is found to increase as well.

Safety Analysis of Riding at Intersection Entrance Using Video Recognition Technology

To study riding safety at intersection entrance, video recognition technology is used to build vehicle-bicycle conflict models based on the Bayesian method. It is analyzed the relationship among the width of non-motorized lanes at the entrance lane of the intersection, the vehicle-bicycle soft isolation form of the entrance lane of intersection, the traffic volume of right-turning motor vehicles and straight-going non-motor vehicles, the speed of right-turning motor vehicles, and straight-going non-motor vehicles, and the conflict between right-turning motor vehicles and straight-going non-motor vehicles. Due to the traditional statistical methods, to overcome the discreteness of vehicle-bicycle conflict data and the differences of influencing factors, the Bayesian random effect Poisson-log-normal model and random effect negative binomial regression model are established. The results show that the random effect Poisson-log-normal model is better than the negative binomial distribution of random effects; The width of non-motorized lanes, the form of vehicle-bicycle soft isolation, the traffic volume of right-turning motor vehicles, and the coefficients of straight traffic volume obey a normal distribution. Among them, the type of vehicle-bicycle soft isolation facilities and the vehicle-bicycle traffic volumes are significantly positively correlated with the number of vehicle-bicycle conflicts. The width of non-motorized lanes is significantly negatively correlated with the number of vehicle-bicycle conflicts. Peak periods and flat periods, the average speed of right-turning motor vehicles, and the average speed of straight-going non-motor vehicles have no significant influence on the number of vehicle-bicycle conflicts.

Factors influencing the usage of shared E-scooters in Chicago

The rapid popularity growth of shared e-scooters creates the necessity of understanding the determinants of shared e-scooter usage. This paper estimates the impacts of temporal variables (weather data, weekday/weekend, and gasoline prices) and time-invariant variables (socio-demographic, built environment, and neighborhood characteristics) on the shared e-scooter demand by using four months (June 2019- October 2019) period of data from the shared e-scooter pilot program in Chicago. The study employs a random-effects negative binomial (RENB) model that effectively models shared e-scooter trip origin and destination count data with over-dispersion while capturing serial autocorrelation in the data. Results of temporal variables indicate that shared e-scooter demand is higher on days when the average temperature is higher, wind speed is lower, there is less precipitation (rain), weekly gasoline prices are higher, and during the weekend. Results related to time-invariant variables indicate that densely populated areas with higher median income, mixed land use, more parks and open spaces, public bike-sharing stations, higher parking rates, and fewer crime rates generate a higher number of e-scooter trips. Moreover, census tracts with a higher number of zero-car households and workers commuting by public transit generate more shared e-scooter trips. On the other hand, results reveal mixed relationships between shared e-scooter demand and public transportation supply variables. This study's findings will help planners and policymakers make decisions and policies related to shared e-scooter services.

Assessment of Crash Occurrence Using Historical Crash Data and a Random Effect Negative Binomial Model: A Case Study for a Rural State

This work identifies factors that influence crash occurrence within a traffic analysis zone (TAZ) by accounting for location-specific effects and serial correlation in longitudinal crash data. This is accomplished by applying a random effect negative binomial (RENB) model. Unlike commonly used count models such as Poisson and negative binomial (NB), RENB accounts for heterogeneity and serial correlation in crash occurrence. An RENB was applied to 15 years of crash data in Arkansas with 1,817 TAZs. Four models were developed for total crashes and by severity (property damage only (PDO), injury, and fatal). RENB-estimated impacts were measured using the incidence rate ratio (IRR). The significant causal factors found to increase in observed crashes include: (i) average precipitation (a one-unit increase in average precipitation results in a 134% increase in total monthly crashes for a TAZ); (ii) average wind speed (16%); (iii) urban designation (7%); (iv) traffic volume (2%); and (v) total roadway mileage (1% for each functional class). Snow depth and days of sunshine were found to decrease the number of accidents by 15% and 2%, respectively. Employment and total population had no impact on crash occurrence. Goodness-of-fit comparisons show that RENB provides the best fit among Poisson and NB formulations. All four model diagnostics confirm the presence of over-dispersion and serial correlation indicating the necessity of RENB model estimation. The main contribution of this work is the identification of crash causal factors at the TAZ level for longitudinal data, which supports data-driven performance measurement requirements of recent federal legislation.

Identifying Temporal Aggregation Effect on Crash-Frequency Modeling

Aggregation of spatiotemporal data can encounter potential information loss or distort attributes via individual observation, which would influence modeling results and lead to an erroneous inference, named the ecological fallacy. Therefore, deciding spatial and temporal resolution is a fundamental consideration in a spatiotemporal analysis. The modifiable temporal unit problem (MTUP) occurs when using data that is temporally aggregated. While consideration of the spatial dimension has been increasingly studied, the counterpart, a temporal unit, is rarely considered, particularly in the traffic safety modeling field. The purpose of this research is to identify the MTUP effect in crash-frequency modeling using data with various temporal scales. A sensitivity analysis framework is adopted with four negative binomial regression models and four random effect negative binomial models having yearly, quarterly, monthly, and weekly temporal units. As the different temporal unit was applied, the result of the model estimation also changed in terms of the mean and significance of the parameter estimates. Increasing temporal correlation due to using the small temporal unit can be handled with the random effect models.

Statistical models for analyzing count data: predictors of length of stay among HIV patients in Portugal using a multilevel model

BackgroundThis study offers a comprehensive approach to precisely analyze the complexly distributed length of stay among HIV admissions in Portugal.ObjectiveTo provide an illustration of statistical techniques for analysing count data using longitudinal predictors of length of stay among HIV hospitalizations in Portugal.MethodRegistered discharges in the Portuguese National Health Service (NHS) facilities Between January 2009 and December 2017, a total of 26,505 classified under Major Diagnostic Category (MDC) created for patients with HIV infection, with HIV/AIDS as a main or secondary cause of admission, were used to predict length of stay among HIV hospitalizations in Portugal. Several strategies were applied to select the best count fit model that includes the Poisson regression model, zero-inflated Poisson, the negative binomial regression model, and zero-inflated negative binomial regression model. A random hospital effects term has been incorporated into the negative binomial model to examine the dependence between observations within the same hospital. A multivariable analysis has been performed to assess the effect of covariates on length of stay.ResultsThe median length of stay in our study was 11 days (interquartile range: 6–22). Statistical comparisons among the count models revealed that the random-effects negative binomial models provided the best fit with observed data. Admissions among males or admissions associated with TB infection, pneumocystis, cytomegalovirus, candidiasis, toxoplasmosis, or mycobacterium disease exhibit a highly significant increase in length of stay. Perfect trends were observed in which a higher number of diagnoses or procedures lead to significantly higher length of stay. The random-effects term included in our model and refers to unexplained factors specific to each hospital revealed obvious differences in quality among the hospitals included in our study.ConclusionsThis study provides a comprehensive approach to address unique problems associated with the prediction of length of stay among HIV patients in Portugal.

Gasoline prices and their relationship to the number of fatal crashes on U.S. roads

Economic factors are assumed to have substantial effects on transportation crash trends. This study makes a comprehensive examination of the relationship between the retail gasoline price, and fatal traffic crashes from 2007 to 2016 in the U.S. Random effect negative binomial (RENB) regression models are used to investigate the relationship while controlling for the other effective factors. This is achieved by developing appropriate crash prediction models. Based on the role of gender and rider's transportation mode in interpreting the prediction results, seven models are developed to investigate the relationship between the inflation adjusted gasoline price and the number of fatal crashes in male, female, motorcyclist, non-motorcyclist, bicyclist, pedestrian, and total groups. Our findings suggest that increasing gasoline prices will not significantly alter the number of total fatal crashes. On the other hand, it is estimated that a one-dollar increase in adjusted gasoline price is associated with a 24.4% increase in the number of fatal motorcycle crashes (p-value = 3.7e-14), a 1.9% decrease in the number of fatal non-motorcycle crashes (p-value = 0.02), and a 8% decrease in the number of fatal pedestrian crashes (p-value = 0.08). There is no noticeable difference between males and females in response to the gasoline price changes.

A crash risk identification method for freeway segments with horizontal curvature based on real-time vehicle kinetic response

With the development and maturation of vehicle-based data acquisition technology, in-vehicle data is increasingly being used to explore road safety. This paper reports on research that analyzed the real-time tire force data (kinetic response) obtained from vehicle kinetic experiments, and constructed a new approach for identifying the high-risk of crashes on freeway segments with horizontal curvature. First, the road was divided into 1km units. Then, taking into account the characteristics of freeway alignment, each segment with horizontal curve was selected as the object of subsequent analysis. Automotive instrumentation was used to obtain a measure of tire force in the course of normal driving. The entire data set was preprocessed according to rate of change and the density of the data was reduced. By defining the outliers of the kinetic data and conducting factor analysis, two representative crash risk indicators of longitudinal and lateral stability were obtained. Negative binomial regression model (NBR model) and random effects negative binomial regression model (RENBR model) were constructed and jointly applied based on the new indicators to predict the risk value of horizontal curve segments. The method showed good prediction performance (71.8 %) for high-risk road segments with design flaws, but the predicted effect for low-risk road segments was not ideal. This study not only illustrated the effectiveness of in-vehicle data in assessing road crash risk by coupling multiple kinetic parameters, but also provided support for freeway safety research using surrogate measures of risk when there is a lack of crash statistics.

Do Institutional and Macroeconomic Factors Matter in IT companies M&As? Evidence from India

The liberalisation of the Indian economy, along with targeted policies of the Indian government for the information technology (IT) sector, has led to tremendous growth in this sector. While there has been substantial foreign investment in India in the IT sector through setting up of subsidiaries, joint ventures and acquisitions, Indian firms have also aggressively invested within and outside India through multiple modes. Among different sectors, the IT sector has been the most dominant in terms of mergers and acquisitions (M&As). This article focuses on host country (macroeconomic and institutional) determinants of M&As undertaken by Indian IT firms. Random effect negative binomial model in panel set up was selected to estimate the models. Four models were estimated as per the availability of data for 35–42 countries and for the period 2000–2015. The results for most of the variables in four models were very similar, reflecting robustness of the methodology. Most of the macroeconomic and institutional factors were found to be important determinants of the M&As by Indian IT companies. However, the economic recession of 2008–2009 was found to significantly reduce the M&A activities by Indian IT companies.

Analysis of Vehicle Breakdown Frequency: A Case Study of New South Wales, Australia

Traffic incidents such as crashes, vehicle breakdowns, and hazards impact traffic speeds and induce congestion. Recognizing the factors that influence the frequency of these traffic incidents is helpful in proposing countermeasures. There have been several studies on evaluating crash frequencies. However, research on other incident types is sparse. The main objective of this research is to identify critical variables that affect the number of reported vehicle breakdowns. A traffic incident dataset covering 4.5 years (January 2012 to June 2016) in the Australian state of New South Wales (NSW) was arranged in a panel data format, consisting of monthly reported vehicle breakdowns in 28 SA4s (Statistical Area Level 4) in NSW. The impact of different independent variables on the number of breakdowns reported in each month–SA4 observation is captured using a random-effect negative binomial regression model. The results indicate that increases in population density, the number of registered vehicles, the number of public holidays, average temperature, the percentage of heavy vehicles, and percentage of white-collared jobs in an area increase the number of breakdowns. On the other hand, an increase in the percentage of unrestricted driving licenses and families with children, number of school holidays, and average rainfall decrease the breakdown frequency. The insights offered in this study contribute to a complete picture of the relevant factors that can be used by transport authorities, vehicle manufacturers, sellers, roadside assistance companies, and mechanics to better manage the impact of vehicle breakdowns.

Understanding Social Interaction across Social Network Sites

ABSTRACT People tend to utilize multiple social network sites (SNSs) simultaneously to maintain some social relationships, which results that there are many overlapping relationships and interactions among SNSs. Although many studies have focused on social interaction and cross-SNS user footprint analysis and understanding, little research investigates social interaction from a perspective of two or more SNSs, and the interplay of interaction among SNSs has been unknown. In this paper, we aim to explore whether interaction building in a new SNS hinders the interaction frequency in an existing site, and if so, what kinds of users and relationships’ interactions are more or less likely to be affected. For these questions, we sampled 7,015 pairs of overlapping identities, 23,590 pairs of overlapping relationships and 6,771 pairs of overlapping interactions from Weibo and Douban and made analysis by combining multiple methods like Regression Discontinuity Design and Random-effects Negative Binomial Regression model. Our results suggest that no matter from the perspective of individuals or from the perspective of relationships, interaction construction in a new SNS is detrimental to interaction frequency in an existing site. Based on our findings, we also propose several valuable insights about how to enhance social interaction and promote its retention when users are involved into interacting practice in multiple platforms.