ISSN 0439-755X
CN 11-1911/B


    26 November 2017, Volume 49 Issue 11 Previous Issue    Next Issue

    For Selected: Toggle Thumbnails
     Parafoveal preview benefits during silent and oral reading
    GAO Min, LI Lin, XIANG Huiwen, SUI Xue, Ralph Radach
    2017, 49 (11):  1357-1369.  doi: 10.3724/SP.J.1041.2017.01357
    Abstract ( 3457 )   PDF (732KB) ( 87 )  
     The researches pay more attention to the parafoveal preview benefits (PPB) in the field of reading and language comprehension. Most of previous studies about PPB were done in silent reading. Those studies mainly investigated whether there is PPB and what type the PPB is. Seldom researches are done about oral reading. We cannot find the literature about the extraction time of preview information during Chinese reading. Thus, present study used the boundary paradigm to explore the differences of PPB between silent and oral reading in Chinese. In experiment 1, in order to explore the effect of the preview of word N when word N-1 was fixated during silent and oral reading, we manipulated the preview condition of word N (masking preview and target preview) and reading modes (silent reading and oral reading). The results show that parafoveal preview information plays an important role in eye movement control during oral reading and silent reading. The preview benefits in oral reading are smaller than that in silent reading. In experiment 2, to explore the effect of the extraction time of preview information of word N when word N-1 was fixated during silent and oral reading, we manipulated parafoveal preview time (0 ms, 50 ms, 100 ms and 150 ms) and reading modes (silent reading and oral reading). The results show that during silent reading, the extraction of parafoveal information might start immediately at the beginning of fixating the word N, but during oral reading, it might start relatively late. In summary, during Chinese reading, parafoveal preview can extract the information during silent reading and oral reading. However, there is significant difference of parafoveal process between silent reading and oral reading. First of all, PPB in silent reading is bigger than that in oral reading. Secondly, parafoveal preview information extraction during oral reading is later than that during silent reading. The last but not least, parafoveal information extraction did not start immediately when pre-target word was fixated. This finding is consistent with the theory of sequential attention shift. But there is not the same phenomenon in silent reading.
    Related Articles | Metrics
     Right hemispheric dominance in forming novel semantic associations
    ZHAO Qingbai, WEI Linlin, LI Ying, ZHOU Zhijin, ZHAO Lili, TANG Lei
    2017, 49 (11):  1370-1382.  doi: 10.3724/SP.J.1041.2017.01370
    Abstract ( 1691 )   PDF (1678KB) ( 118 )  
     Forming novel associations is one of the key subprocesses in creative thinking. According to the generation-selection model of creativity and the theory of coarse semantic coding in right hemisphere, the right brain may play a dominating role in forming novel associations. However, this inference is only supported by half of the neuroimaging studies, and the others emphasize the role of left hemisphere or both. The disagreement among previous studies might result from different materials and different task requirements they used. The right hemispheric dominance in forming novel semantic associations may be more easily detected in the tasks of comprehending creative language. In the current study, the Chinese two-part allegorical sayings as well as fMRI and ERPs were used to explore the temporal and spatial features of neural processing in forming novel semantic associations. Two experiments were conducted in this study. In Experiment 1, participants were asked to read silently some Chinese two-part allegorical sayings and keep them in mind. For half of the experimental materials, the semantic association between the two parts was novel, while for the other half, it was normal. The neural activity was recorded by fMRI. In Experiment 2, the experimental materials were similar with those in Experiment 1, except for some filling materials in which there was not semantic association between the two parts. Participants were asked to judge whether the first part was associated with the second part in meaning, and the neural activity was recorded by ERP. The result of Experiment 1 showed that compared with normal semantic association, novel semantic association activated more in right superior temporal gyrus, which was related to the retrieval of novel semantic information. Results of Experiment 2 showed that novel semantic association induced a more positive late positive component (650~900 ms after the onset of target words) over the right temporal sites and right frontal sites, which might reflect the retrieval as well as selection and integration of novel semantic information respectively. In summary, the current results support the right hemispheric dominance in forming novel semantic associations, and indicate that its timing may be at the late period of processing.
    Related Articles | Metrics
     Deaf students’ syntactic awareness of the “shi…de” construction in sentence reading
    ZHANG Fan, LI Degao
    2017, 49 (11):  1383-1391.  doi: 10.3724/SP.J.1041.2017.01383
    Abstract ( 1375 )  
     Deaf students take sign language and written language as their first and second language, respectively. Probably because of differences between the two languages in modality, they meet serious problems with grammar in written language. However, much remains unknown behind this phenomenon. In China students with a profound level of hearing loss (SPL) attend the same activities on campus as those who have a moderate or slight level of hearing loss (SM/SL). Since most SM/SLs are able to use spoken Chinese with the help of hearing aids but SPLs are not, however, it is likely that the former develop a better awareness of grammar in written Chinese than the latter. We conducted three experiments on a cohort of 117 deaf college students (55 SPLs) from Zhejiang Vocational College of Special Education, China with sentences of the “shi…de” construction as the materials in a moving-window self-paced reading task. The structural particles “shi” and “de” in a “shi…de” sentence form a syntactic construction so that what is in between is pragmatically emphasized. The words between the two particles were an adjective, verb-object phrase, and verb phrase modified by a prepositional phrase in Experiment 1, 2, and 3, respectively. In each experiment, the key sentences were common examples of a “shi…de” construction and the corresponding grammatically incorrect sentences. A correct sentence was converted into an incorrect sentence with the first particle “shi” removed. There were 36, 34, and 36 pairs of critical sentences in Experiment 1, 2, and 3, respectively. It was hypothesized that a stronger wrap-up effect would be revealed in the participants’ reaction times to the second particle “de” in the incorrect than in the correct sentences. The wrap-up effect means that readers spend more time to the last word of a sentence because extra resources of cognition are needed for the integration of information. The SM/SLs would be more likely than the SPLs to have a wrap-up effect in the correct sentences. We obtained two main results. (1) As expected, a stronger wrap-up effect was observed in the SM/SLs’ reading time to “de” in the incorrect than in the correct sentences. However, there was no significant difference in the SPLs’ reading time to the particle between the types of sentences in Experiment 1 and 3. (2) The SPLs had significantly longer reading times to “de” in the correct than in the incorrect sentences in Experiment 2. It was concluded that the SPLs developed a significantly weaker awareness of grammar than the SM/SLs with the “shi…de” construction. More studies are needed to reveal deaf students’ syntactic awareness in written language in general.
    Related Articles | Metrics
    Effects of “each speaks their own dialect” phenomenon on the executive function of Jingpo students
    WANG Ting, WANG Dan, ZHANG Jijia, CUI Jianai
    2017, 49 (11):  1392-1403.  doi: 10.3724/SP.J.1041.2017.01392
    Abstract ( 1773 )  
     Many Jingpo families are composed of people from different branches of the Jingpo nationality, each of which has its own dialect. Family members use these different dialects when they communicate with each other. Will this linguistic phenomenon influence their executive function? Previous research on the relation of language and executive function views inhibitory control as one of the common components but seldom takes into consideration the diversity of executive functions. Moreover, previous studies have confirmed that bilingualism has a positive effect on many cognitive functions. Studies of cognitive psychology and cognitive neuroscience have strongly suggested that executive function is not a unitary construct. However, few researchers have examined the ethnic Chinese minorities through empirical study The present study investigated three latitude executive functions (inhibition of responses, memory updating, and attention switching) of 56 Jingpo college students in Yunnan Province. The participants of this study were divided into groups that do and do not speak their own dialect. Both groups were investigated in three experiment tasks to explore the effects of language experience and culture on different components of executive function. In Test 1, the inhibition subcomponent was measured by the Stroop task and stop–signal task. In Test 2, the updating subcomponent was tested by the active memory task and color dots updating task. In Test 3, the shifting subcomponent was tested by local–global task and more–odd task. Correlation analysis results indicated that tasks measuring the same executive function presented a significant positive correlation, and the correlation between tasks measuring different functions was not significant, providing strong evidence for the diversity of executive function. Based on this result we tested the cognitive advantages in three executive functions between students who do and do not speak their own dialect. The result indicated a bilingual cognitive advantage in Jingpo college students. Students who speak their own dialect outperformed those who do not in inhibitory control and shifting, but there were no significant difference of two groups in memory updating. In summary, the present study suggests that executive function includes different components, and that language experience has a specific effect on executive function. These can be new evidence for the hypothesis of linguistic reality: language influence cognition.
    Related Articles | Metrics
     The effect of tennis expertise on motion-in-deep perception: An event-related potential study
    WEI Xiaona, QI Changzhu, XU Xia, HONG Xiaobin, LUO Yuejia
    2017, 49 (11):  1404-1413.  doi: 10.3724/SP.J.1041.2017.01404
    Abstract ( 1720 )  
     Perception of motion-in-depth is essential to guide and modify drive action in interceptive-dominated sports. Tennis players usually exhibit better performance in perception of motion-in-depth. There is considerable evidence for cognitive advantages of sports expertise, but the neural mechanisms of these benefits remain poorly understand. The purpose of this study is to investigate the effect of motor expertise on perception of motion-in-depth by event-related potential (ERP) with expert-novice paradigm. Two groups of volunteers, one comprising 19 expert tennis players with skill level 2 (trained for at least 5 years) and the other 19 novices, participated in the experiment. There were six videos made as the stimuli of the experiment, including four target stimuli (a directly-approaching sphere, a directly-receding sphere, a rotating approaching sphere and a rotating receding sphere) and two non-target stimuli (an expanding circle plane and a contracting circle plane). In each trial, participants watched one video and were asked to identify the target and its motion direction as quickly as they could (if the participant didn’t response in one second, the video ended and it would be considered as a wrong answer). Their reaction time, accuracy and EEG data were recording synchronously when participants completed the cognitive task of motion-in-depth perception. There were a total of 360 trials divided into three blocks of 120 trials each. Each block included 20 repetitions of the six stimuli. ERP data of five electrode sites was selected for further statistical analysis: PO7 and PO8 for P1 and N180, Oz for P2, CPz and Pz for P300. Peak amplitudes (baseline - peak measured) and corresponding latencies were detected in four windows following stimulus onset: 80~160 ms (P1), 160~220 ms (N180), 220~280 ms (P2), and 320~580 ms (P300). Experimental design with 2 factors (direction of motion in depth: approaching vs. away) × 2 factors (rotation pattern: rotating vs. non-rotating) were adopted and further repeated measures of ANOVA in amplitudes and latencies of these ERP components were conducted in group statistical analysis. The results indicated that there was similar processing dynamics in ERP data between expert tennis players and novices. We also detected three main differences between experts and novices: 1) the accuracy of motion-in- deep perception in experts was significantly higher than that of novices; 2) the latency of P1 was significantly longer in “approaching” condition than “away” condition in novices, but experts didn’t show any differences in these two conditions; 3) the latency of P2 was significantly longer in “approaching” condition than in “away” condition in experts, however, signi?cant differences were not found in novices. These results indicated that experts’ better performance in accuracy in perception of motion-in-deep is possibly related to selective attention and pattern recognition. Shorter latency of P2 might be a potential marker to evaluate the ability of perception of motion-in-deep. Additionally, this study will further deepen our understanding for the neural mechanisms of sport expertise.
    Related Articles | Metrics
     Bitterness followed by happiness: A fMRI study on English lovers
    LIU Lulu, XIAO Jing, ZHOU Jianshe, LU Jiamei, LUO Jing
    2017, 49 (11):  1414-1427.  doi: 10.3724/SP.J.1041.2017.01414
    Abstract ( 2825 )  
     Academic emotions play important roles in academic motivation and achievement. However, most of the studies on academic emotions focused on negative ones such as test anxiety, the positive academic emotions such as pride, enjoyment, and hope are far less investigated, let alone the brain basis underlying them. In this study, we explored the neural correlates of positive academic emotions by functional magnetic resonance imaging (fMRI). College students who have high positive academic emotions for learning English (English lovers) were selected as participants. They were required to remember and recognize English or Chinese materials while their brain activation were recorded. Through comparing the neural correlates involved in processing the Chinese and English materials in different stages, which critically included the learning and testing stages, the mental preparation stage before learning and testing, and the feedback of memory performance stage, we can identify the cognitive brain processes, especially the emotional and motivational ones, characterized English lovers. English lovers were selected by English Happy-Learning Questionnaire (EHQ), together with Maslach Burnout Inventory-General Survey (MBI-GS), Self-Rating Anxiety Scale (SAS), and Self-Rating Depression Scale (SDS) to eliminate possible confounders. Nineteen college students (nine males and ten females) who meet the criterion of the study participated in this experiment. Unfamiliar city-state name pairs, which could be presented in English (e.g., Conakry-Guinea) or in Chinese (e.g., 阿塔富-托克劳), were used as materials. Participants were instructed to remember 144 city-state name pairs (72 for each language), and then they were given the memory (recognition) test. We focused not only on the brain activation involved in memorizing and recognizing English/Chinese materials, but also on those involved in the mental preparation for learning and testing, and that in getting positive or negative feedbacks in memory performance. Our results revealed two major points: (1) In the learning stage, more anterior insula activation were found when the “English lovers” were making mental preparation for learning English relative to Chinese, and this difference in insular activation were found to be positively correlated with the scores of EHQ and that of performance in memory test, implying “English lovers” took more pain and mobilized more cognitive effort in making preparation for learning English. (2) In the retrieval and feedback stage, more activation in the regions for mental reward (midbrain and substantia nigra) and emotional arousal (amygdala) were found to be associated with the receiving of positive feedback for English than for Chinese, whereas more activation in the pain and disgust region (insula) were associated with the receiving of negative feedback for English than for Chinese, both kinds of brain activation differences were positively correlated with the scores of EHQ, implying the success or failure in learning English can have more emotional and motivational implications than its experimental equivalents in Chinese. Taken together, the present study indicated that the emotional, motivational, and cognitive brain processes of the positive academic emotions for learning English mainly embodied in the stages for making mental preparation for learning or testing and for processing feedback on memory performance, but not in the actual learning and testing processes. To be fond of learning may not mean a kind of pure enjoyment, rather it is something that started with bitterness and ended with sweetness.
    Related Articles | Metrics
     Longitudinal linkages between social support, posttraumatic stress disorder, and posttraumatic growth among primary school students after the Ya’ an earthquake
    ZHOU Xiao, WU Xinchun, WANG Wenchao, TIAN Yuxin
    2017, 49 (11):  1428-1438.  doi: 10.3724/SP.J.1041.2017.01428
    Abstract ( 2961 )   PDF (464KB) ( 256 )  
     It has been repeatedly documented that posttraumatic stress disorder (PTSD) is a common negative outcome in trauma-related research. Nevertheless, traumatized individuals also reported positive psychological changes in various aspects of life such as personal strength, relationship, and appreciation of life, that were collectively defined as posttraumatic growth (PTG). PTG has been described in many studies as evident in survivor reports following a wide range of traumatic events. Importantly, PTSD and PTG may co-exist in trauma survivors, leading to a heated discussion on the relation between PTSD and PTG. Nevertheless, the relation between PTSD and PTG still remains mixed. For example, some studies found a positive relation between them, indicating that individuals who had high PTSD reported increased PTG. In contrary, some found a negative relation, suggesting that PTG and PTSD could be conceptualized as the two ends of the same continuum. Some even failed to find a significant relation between them, arguing that PTSD and PTG were two independent constructs. The inconsistent findings inhibited our understanding in the relation between the two constructs, thus it is necessary to examine simultaneously the predictive factors of PTSD and PTG to elucidate their relation. Wherein, social support is considered as one of the most important predictors of posttraumatic reactions. Several theories were proposed to explain the role of social support in posttraumatic outcomes, and studies indicated that social support was a protector for PTSD while a facilitator for PTG. Nevertheless, few research has examined the linkages between social support, PTSD, and PTG from a longitudinal perspective. In order to advance the gaps and to assess the longitudinal relation between social support, PTSD, and PTG, we selected 303 children to fill out a social support questionnaire, a children posttraumatic stress scale, and a posttraumatic growth inventory at 6 months (T1), 12 months (T2), and 18 months (T3) after the Ya’an earthquake, respectively. Firstly, a cross-lagged model between PTSD and PTG was built, and the results indicated that PTG had no significant effect on PTSD from T1 to T3, but PTSD at T1 had a significant positive effect on PTG at T2. However, the effect of PTSD at T2 on PTG at T3 was non-significant. Then, we inserted social support into the above cross-lagged model, and found no change in the relation between PTSD and PTG. Additionally, social support at T1 had a direct and negative effect on PTSD at T2 but not vice-versa, and there were not any significant mutual effects between social support and PTG from T1 to T2. Moreover, PTSD at T2 had a significant negative effect on social support at T3 but not vice-versa, and there were significant positive and mutual effects between social support and PTG from T2 to T3.
    Related Articles | Metrics
     What do you listen to under the pressure of time? The moderator effects of reference group on impulsive buying
    ZHOU Yuanyuan, HU Yangli, ZHAO Yancheng
    2017, 49 (11):  1439-1448.  doi: 10.3724/SP.J.1041.2017.01439
    Abstract ( 3902 )  
     Sometimes consumers should make decisions under the pressure of time. Although researchers have investigated impulsive buying under time pressure for many years, almost no consensus has reached. A number of literature have found that time pressure has a positive impact on impulsive buying while other literature have got the opposite conclusion. These researches ignored others’ influence. With the popularization of the social media, consumers are more and more exposed to reference group when making decision under the pressure of time. So in this research we will investigate the relationship between time pressure and impulsive buying under different types of reference group influence based on reference group theory. Three studies were conducted to verify our hypotheses. Study 1 used secondary data and an experiment to test main effect. First, 104 products’ sale data from JUMEI have been gotten. Based on regression analysis, we verified the interaction effects of time pressure and reference group on impulsive buying. Then, experiment 1 was conducted through a 2 (time pressure: high vs. low) × 2 (reference group: informational influence vs. normative influence) between-subjects design. 128 participants completed the experiment. The results also revealed the same significant interaction effects. Study 2 used the similar experimental design as study 1 to further test our mediation effects. 148 university students from Zhongnan University of Economics and Law participated in the experiment. The results revealed the significant mediation effects of instant gratification (i.e., the degree of immediate gratification that an individual obtains via making an impulsive buying) and normative evaluation (i.e., judgments about the appropriateness of engaging in impulsive buying behavior). Moreover, study 3 further tested informational influence. A total of 144 participants from Huazhong University of Science and Technology completed the study with a 2 (time pressure: high vs. low) × 2 (informational influence: content information vs. quantity information) between-subjects design. The results revealed the significant interaction effects of time pressure and informational influence on impulsive buying. The results of these three studies have provided supports for our hypotheses: (1) The reference group moderate the relationship between time pressure and impulsive buying—Specifically, consumers under higher time pressure will do more impulsive buying in the normative reference influence, while those under lower time pressure will do more impulsive buying in the informational reference influence. (2) Instant gratification and normative evaluation mediate the above relationships. (3) The informational influence moderate the relationship between time pressure and impulsive buying—Consumers under lower time pressure will do more impulsive buying when reading content information, while those under higher time pressure will do more impulsive buying when reading quantity information. Finally, we discussed the theoretical contributions and managerial implications of this paper, and offered some critical insights for marketers.
    Related Articles | Metrics
     Resolving “Commuting Paradox”: How commute time influences subjective well-being
    WU Weijiong
    2017, 49 (11):  1449-1459.  doi: 10.3724/SP.J.1041.2017.01449
    Abstract ( 3568 )  
     People in today’s society spend a substantial amount of their time traveling to and from work. Researchers have rightfully concerned themselves with the question of if and how commuting affects people’s lives. Some behavioral economists suggested that commute time play a negative effect on individuals’ life satisfaction. This phenomenon is called “commuting paradox”, in which individuals’ utility are imbalance due to longer commuting time is not compensated. The present study regards commute time as the work-family transition zone, such as social transition zone and psychological transition zone. With these perspectives, we aimed to examine the moderating roles that marital status (social transition zone) and recovery experiences (psychological transition zone) play in the relationship between commute time and subjective well-being. What is more, the mediating mechanism of commuting utility was explored. In order to test our model, we conducted a survey on 822 part-time graduates from three colleges. Data were collected from 3 follow-up surveys to avoid the common method bias. Participants were asked to fill out questionnaires at three time points (Time 1: commute time, marital status and recovery experiences; Time 2: commuting utility; and Time 3: satisfaction with life, happiness). These variables were assessed by: commute time survey, marital status survey, recovery experiences questionnaire, satisfaction with life scale, PANA scale, and Princeton affect and time survey. All Cronbach’s alpha coefficients were acceptable (ranging from 0.83 to 0.91). Descriptive statistics and hierarchical polynomial regression analysis were applied to test the hypotheses. The results indicated that: (1) marital status (social transition zone) moderated effects of commute time on subjective well-being, i.e., unmarried employees’ commute time had negative impact on life satisfaction, married employees’ commuting time had U shape impact on life satisfaction, happiness and occupational well-being; (2) recovery experiences during work→home commute (psychological transition zone) moderated effects of commute time on outcome variables, i.e., psychological detachment moderated relationships between unmarried employees’ commute time and commuting utility; relax experience moderated the relationship between unmarried employees’ commute time and happiness; (3) effects of married employees’ commute time on commuting utility and happiness were moderated by relax experience, whereas the relationship between married employees’ commute time and life satisfaction were moderated by psychological detachment; (4) commuting utility not only mediated the effects of commute time on life satisfaction and happiness, but also mediated the moderations of marital status and recover experiences; (5) employees’ utility equilibrium were found during “commuting time trap” (1.75 h - 2.75 h), in which longer commuting time was compensated. Significance: The present study analyzed the commuting paradox from two aspects, including social transition zone and psychological transition zone. Then we built a theoretical model regarding how commute time influences employee’s subjective well-being. Together, our findings contribute to the literature by helping to (a) provided a psychological explanation for commuting paradox, (b) integrate commuting utility, life satisfaction and happiness, (c) resolve mixed findings regarding the issue of commute time and subjective well-being. The managerial implications of our findings, limitations, as well as future research directions were discussed.
    Related Articles | Metrics
     Priming effects of virtual avatars on aggression: Influence of violence and player gender
    HENG Shupeng, ZHOU Zongkui, NIU Gengfeng, LIU Qingqi
    2017, 49 (11):  1460-1472.  doi: 10.3724/SP.J.1041.2017.01460
    Abstract ( 4809 )   PDF (613KB) ( 163 )  
     A virtual avatar is a video game player’s self-presentation in virtual space. The physical appearance of an avatar can prime stereotypes and behavioral scripts stored in memory. The relation between avatar appearance and aggression has been substantially confirmed, but there are open questions about the conditions in which this association is strongest, and what the relation between avatar identification and aggression is. This study used a cue-priming paradigm in two experiments to test the effect of avatar appearance on avatar identification and aggression in violent and nonviolent video games; to test gender as a moderator of these effects; and to test the correlation between avatar identification and aggression. The first experiment investigated the effect of avatar appearance on the level of avatar identification and aggression in violent and nonviolent video games. This experiment employed a 2 (Avatar Appearance: justice/evil) × 2 (Game Violence: violent/non-violent) between-subjects design. 75 male participants were randomly assigned to play a violent or non-violent video game using an avatar representing justice or evil. Based on the first experiment, the second experiment explored the interaction effect of avatar appearance and gender on the level of avatar identification and aggression in a violent video game. This experiment also employed a 2 (Avatar Appearance: justice/evil) × 2 (Gender: male/female) between- subjects design. 42 male and 36 female participants were randomly assigned to play a violent video game using an avatar representing justice or evil. After game play, the amount of hot sauce given by participants to an ostensible partner who hated spicy food was used to measure aggression, and an avatar identification scale was used to measure identification with the avatar. The results showed that: (1) The relations between avatar appearance, avatar identification and aggression were influenced by the violence of game. In the violent video game, the identification with the justice avatar was significantly higher than with the evil avatar, and the evil avatar elicited significantly higher aggression than the justice avatar. In the non-violent game, there was a marginally significant difference between the identification with the justice or evil avatar, but there was no significant difference between the level of aggression elicited by the justice or evil avatar. (2) In the violent video game, the avatar effects were moderated by player gender. Specifically, the avatar identification of female participants was significantly affected by avatar appearance, whereas that of the male participant was not. Avatar appearance had a stronger impact on the aggression of males than females. (3) There was a significant correlation between avatar identification and aggression, which was moderated by game violence and gender. In conclusion, the results of this study supported the priming effect theory and were partially consistent with the existing research. Several factors influenced the effect of avatar appearance on aggression, including a video game factor (violent or non-violent game) as well as an individual factor (male or female), and the complex relationship between avatar identification and aggression. One social implication of the study is that game designers should embed more positive associations, situations, and stereotypes in games to provide users more positive potential priming effects.
    Related Articles | Metrics
     Performance of the entropy as an index of classification accuracy in latent profile analysis: A Monte Carlo simulation study
    WANG Meng-Cheng, DENG Qiaowen, BI Xiangyang, YE Haosheng, YANG Wendeng
    2017, 49 (11):  1473-1482.  doi: 10.3724/SP.J.1041.2017.01473
    Abstract ( 2153 )  
     Latent Profile Analysis (LPA) is a latent variable modeling technique that identifies latent (unobserved) subgroups of individuals within a population based on continuous indicators. LPA has become a popular statistical method for modelling unobserved population heterogeneity in social and behavioral science. Entropy is a standardized index of model-based classification accuracy, with higher values indicating more precise assignment of individuals to latent profiles. In lots of conditions, the aim of substantial research was to assign individual to different latent subgroup. Therefore, Entropy was chosen to report as an index reflecting accuracy of class membership assignment. Unfortunately, very few methodological studies have examined the behavior of Entropy under the conditions where sample sizes, latent class separations, number of indicators, and number of classes are varying. Thus, the primary purpose of this study was to examine how Entropy will perform with different sample sizes, latent class separations, number of indicators, and number of classes. By using Monte Carlo simulation techniques, we generated artificial data to fit true models and evaluated the performance of Entropy and entropy-based indexes (CLC, ICL_BIC, sample adjusted ICL_BIC) under different modeling conditions. The simulation was repeated 100 times for each condition of the 120 combinations: sample sizes (50, 100, 500, 1000, 3000), latent class separations (0.5, 1.2, 3), number of indicators (4, 8, 12, 20), and number of latent classes (3, 5). The continuous indicators of the latent class are not allowed to correlate. Different mean levels on the observed variables are calculated by Mahalanobis distance (MD). The simulations and analyses of the sample data were conducted using the Monte Carlo facilities of Mplus7.4. For 3 latent classes, Entropy values round 0.76 and above are related to at least 90% correct assignment, and Entropy values round 0.64 and below are related to at least 20% classification error rate. When the latent classes is 5, Entropy value around 0.84 and above are related to at least 90% correct assignment. The Entropy value decreases and the classification error rate increases as sample size increases. Entropy performs well under small sample sizes (50-100) and more indicators conditions. Entropy consistently performs better when latent class separation is large (MD=3), and the result is quite consistent across the sample size and number of latent classes. The tendency of CLC, ICL_BIC, and sample adjusted ICL_BIC were similar, which increases as sample size increases, and it also increases under large class separation but the differences of Entropy caused by class separation were more noticeable. This simulation indicates that the Entropy values are strongly correlated with the correct class membership assignment, but it varies according to number of latent classes, sample sizes, latent class separation and number of indicators. Hence, it is hard to determine cutoff values for Entropy, the indicator of class assignment.
    Related Articles | Metrics