【作者】 陈贵；

【导师】 蔡太生；

【作者基本信息】 中南大学 ， 心理学， 2013， 博士

【摘要】 目的：延迟贴现任务是一种有效的冲动性行为测量任务,所反映的延迟等待能力与人的非理性决策密切相关。本研究对延迟贴现任务的四种分析模型(两个单参数贴现模型：指数衰减模型和双曲线模型；两个双参数模型：类双曲线模型和幂函数模型)进行比较,对多种测量指标进行意义和适用性分析,旨在进一步优化和统一延迟贴现任务的统计分析方法和测量指标。之后,本研究在统计分析的方法论基础上将延迟贴现任务应用于青少年人群中,以了解超重/肥胖青少年的冲动性特征(除延迟贴现任务外,还包括另一行为任务和另一问卷调查在内所测的冲动性)及其对进食行为的影响。方法：1.研究一,61名普通青少年被试的食物—金钱延迟贴现任务(Delay Discounting Task, DDT)数据用于研究分析。四个贴现模型通过非线性回归的方法对群组和个体贴现数据进行曲线拟合,群组水平通过适配指标(方差变异解释量R2)和拟合曲线比较,个体水平通过可有效拟合的个案数比例、R2的Wilcoxon秩和检验、参数S的t比率统计检验及参数S的Wilcoxon秩和检验,对贴现模型的有效性、适配性、参数S的合理性及所反映的贴现意义进行分析,以挑选最佳贴现模型用于食物—金钱延迟贴现任务的数据拟合。2.研究二,38名正常体重青少年被试(排除其它可能冲动性障碍)完成食物—金钱延迟贴现任务(Delay Discounting Task, DDT)、进食态度量表(Eatting Attitude Test, EAT)和一般人口学资料调查,任务调查完毕后对被试身高和体重进行测量。结合多种贴现指标就它们的贴现意义和适用性进行分析。拟合函数的积分面积(Calculus Area, CA)和曲线下面积(Area-under-the-curve,AUC)之间经配对t检验比较两种方法的计算结果差异。组内食物和金钱贴现面积经配对t检验进行差异性比较(两种方法分别检验),比较两种方法的统计效果差异。挑选CA相近的两组个案贴现数据,结合它们的拟合曲线和参数,加LogK、LogS和CA的相关分析,就积分面积与参数的差异和联系进行分析,具体解释它们所代表的贴现意义。从三贴现指标出发,经配对t检验就组内食物和金钱贴现差异进行比较。3.研究三,从普通青少年人群中抽取52名正常体重青少年(31男,21女)和51名超重/肥胖青少年(30男,21女)完成食物—金钱延迟贴现任务(Delay Discounting Task, DDT)、开门任务(Door Open Task, DOT)、Barratt冲动量表-11(The Barratt impulsiveness Scale-11, BIS)和一般人口学资料的调查。在实验任务测试阶段测量被试身高和体重,用于BMI计算。以人口学因素和超重/肥胖BMI筛查指标对被试进行分组,用类双曲线模型对各实验分组的群组贴现数据进行非线型拟合,并对拟合曲线进行组间比较；用单因素方差分析就人口学变量和体重分组在各冲动性变量上的主效应进行分析；在主效应显著的情况下,用多因素方差分析就人口学因素和实验分组因素之间的交互作用进行分析；采用偏相关方法对各冲动性变量间的相关性进行研究；用相关和回归分析就冲动性变量与BMI之间的关系进行研究。4.研究四,本研究的测量任务、测量程序同于研究三,沿用了研究三中的超重、肥胖者的数据资料,在此基础上继续补充了样本容量,165名超重/肥胖青少年(其中女59人,男84人)的食物—金钱延迟贴现任务(Delay Discounting Task, DDT)、进食态度量表(Eatting Attitude Test, EAT)、进食障碍量表(Eating Disorder Inventory-1,EDI-1)和一般人口学资料数据用于此部分研究。用单因素方差分析就男、女在AUC、进食态度和进食障碍上的差异性进行检验,在考虑性别差异情况下,分性别用皮尔逊相关分析就延迟贴现与进食态度、进食障碍及各分量表之间的相互关系进行分析。采用单因素方差分析就高、低贴现程度组的进食态度、进食障碍差异进行分析；采用单因素方差分析就进食正常组、厌食或贪食倾向组和厌食或贪食症可能组的延迟贴现程度、进食障碍总分及各分量表得分差异性进行分析。结果：1.研究一,在群组和个体水平,R2的比较结果表明双参数模型对贴现数据拟合的适配性要优于单参数模型,两个双参数模型之间并无明显的适配性差异(p>0.05)；单参数模型对部分个案贴现数据不能有效拟合,指数衰减模型有近30%,双曲线模型有近23%,两个双参数模型对所有个案贴现数据均能有效拟合；参数S的t比率统计检验结果表明类双曲线模型有近63%的个案参数S显著小于1,幂指函数模型有近58%的个案参数S显著小于1；幂指函数模型参数S在本研究贴现任务的两种强化刺激物之间并无显著差异(Z=-1.756,p=0.079<0.05),而类双曲线模型的参数S在食物和金钱之间差异明显(Z=-2.817,p=0.005<0.05),可反应被试对奖赏刺激物的敏感性。2.研究二,AUC与CA在具体面积计算结果上存在着显著差异(食物：t(37)=3.073,p<0.005；100元钱：t(37)=4.057,p<0.005),但反映被试平均CA和平均AUC的柱形高低差异不大,两种方法在组内食物和金钱贴现面积差异性比较上统计效果相同(CA: t(37)=2.411,p<0.005; AUC:t(37)=2.480, p<0.005),贴现面积和拟合参数通过贴现函数的微积分计算得到了有机结合。CA大小相近的两个案比较发现拟合参数K可反映曲线的陡峭程度,参数S可反映曲线随延迟时间的水平拉伸情况。CA、LogK和LogS的相关分析发现,CA与LogK无相关(食物：r=0.089,金钱：r=0.055；p>0.05),与LogS有中度负相关(食物：r=-0.413,金钱：r=0.509；p<0.05),LogK与LogS呈高度负相关(食物：r=-0.801,金钱：r=-0.796；p<0.01)。组内食物和金钱延迟贴现任务比较仅发现在CA上差异显著(t=2.411,p<0.05),在LogK和LogS上差异不显著(p>0.05)。3.研究三,本研究中年龄、性别和家庭经济收入三个人口学因素对各冲动性变量的影响不显著(p>0.80)。超重/肥胖组与正常体重组的冲动性差异比较发现他们在AUC和开门尝试次数(N DO)上差异显著,超重/肥胖者在贴现任务中对食物和金钱表现出更高贴现程度(食物：F=15.37,p<0.01；金钱：F=19.28,p<0.01),在开门任务中有更多次数的开门尝试(F=10.74,p<0.01)。超重/肥胖组与正常体重组在其它行为冲动性变量、Barratt冲动量表及各分量表上则没有表现出差异性(p>0.05)。体重分组与人口学因素的交互作用分析发现主效应显著的三个冲动性变量(AUC食物、AUC金钱、N_DO)与性别的交互作用显著(食物：F(1,101)=15.370,p=.000；金钱：F(1,101)=19.283,p=0.000；开门尝试次数：F(1,101)=0.433,p=0.05)。进一步简单效应分析发现,超重/肥胖组和正常体重组的贴现程度差异仅存在于女生中(食物：F=26.38,P<0.01；金钱：F=44.86,P<0.01),而在男生中并不存在(p>0.167)；超重/肥胖组和正常体重组在开门尝试次数上的差异在两性别分组中均存在,但不同性别中两分组之间的差异程度不同(女生：F=4.80,p<0.05；男生：F=5.90,p<0.01)。三种冲动性测量任务之间相关不显著(p>0.05),BMI仅与食物-金钱贴现任务的AUC相关(相关系数分别为-0.229和-0.240),与其它冲动性变量相关不显著(p>0.05)。BMI对AUC具有良好预测作用,特别是在女生群体中,BMI分别能解释食物和金钱AUC26.4%和33.4%方差变异。4.研究四,在男超重/肥胖青少年中,AUC与进食态度、进食障碍及各分量表的相关均不显著(p>0.05)；在女超重/肥胖青少年,AUC与进食障碍的瘦身倾向、贪食、完美主义和内省四个因子表现出适度负相关(person r分别为-0.268、-0.323、-0.297和-0.263)。在超重/肥胖女生中,进食正常组、厌食或贪食倾向组和厌食或贪食症可能组在金钱延迟贴现任务的AUC、进食障碍总分及进食障碍的大多数分量表上差异显著(p<0.05),高、低贴现组仅在贪食和内省两个分量表上表现出得分差异(F=3.752, p<0.05; F=4.098, p<0.05)。结论：1.双参数模型对原始贴现数据的要求低于单参数模型,对贴现数据有效拟合的比例更高,双参数模型非线性拟合的方差变异解释量R2更高,对贴现数据的适配性和描述效度更佳；双参数模型其参数S的增加具有合理性,用类双曲线模型拟合,有更多个案的参数S显著小于1。另外,类双曲线的参数S还可反映被试对延迟强化物的敏感性。所以,用类双曲线模型对食物—金钱延迟贴现任务的数据进行拟合,其适配性和描述效度更佳。2.AUC与CA在实质上均反应了被试在一段延迟时间里的总贴现程度,二者在差异性分析上具有相同统计效度。贴现函数的积分计算将拟合参数和贴现面积有机结合在一起,贴现率K可反映主观价值的变化情况,它与CA并不相关,它们代表了贴现能力的不同方面,此方法也有效地解决了以往贴现面积相等而贴现曲线差异时的描述困难。3.研究结果支持超重/肥胖青少年比正常体重青少年更为冲动的研究假设,但他们之间的冲动性差异仅表现在行为测量任务上,在自我报告的量表任务上两种人群的差异不显著。超重/肥胖青少年对延迟奖赏的贴现程度要显著高于正常体重青少年,在开门任务中尝试了更多次数的开门点击,即表现出即刻满足、对奖赏敏感的冲动性决策缺陷。两种人群的冲动性差异与性别交互作用显著,贴现程度差异在男生群体未达显著,开门尝试次数差异在女生群体中更为显著。三种测量任务反映了冲动性的不同维度,其中延迟贴现与BMI关系更为密切,BMI对贴现程度具有良好预测效度,即BMI值越大(越肥胖),越倾向于选择即时奖赏。4.延迟贴现程度与进食问题(由EAT和EDI反映)之间的关系存在着性别差异,在女超重/肥胖青少年中两方面关系较为密切,在男超重/肥胖青少年中二者相关并不显著。女超重/肥胖青少年的延迟贴现程度与其贪食行为(由EDI反映)相关显著,即越倾向于即时满足,越表现得贪食,另外,女超重/肥胖青少年的延迟贴现程度也与瘦身倾向、完美主义和内省三方面心理特征相关密切,延迟贴现程度越高,三个因子维度上的进食心理问题越明显。差异性分析发现高贴现女超重/肥胖青少年比低贴现对照组表现出严重贪食症状和较低水平的情感内省。另外,厌食或贪食症可能组的超重/肥胖女生比进食正常组的超重/肥胖女生表现出更高贴现程度,进食问题也更为严重。更多还原

【Abstract】 Objective:The Delay Discounting Task (DDT), as an effective behavior measure, reflects the impulsivity characteristic which closely related to human’s irrational decision making. In order to get the statistical analysis method and the measure indexes of DDT optimized and normalized, four prominent models of delay discounting (two one-parameter models:the exponential decay model and the hyperbola model; two two-parameter models:the hyperboloid-like model and the power exponential function model) were compared and the significance and applicability of DDT’s measure indexes were analyzed in our study. Afterwards, DDT was applied in adolescent on the bases of methodology about statistical analysis for a better understanding of the impulsive characteristics (except DDT, also included other impulsivity dimensions be reflected by another behavior measure and a questionnaire) related to adolescents’ overweight/obesity problems and its influence on eating behaviors.Methods:In study1, a sample of61general adolescents completed the food-money related DDT. All four discounting models were fit to the group and individual discounting data using nonlinear regression techniques. The R2values and the fitting curves were direct compared among four models at group level. In the individual level, the effectiveness, suitability and rationality of the four models were also analyzed through results which come form the percentage of effectively fitting cases, the R2values were subjected to Wilcoxon signed-rank tests, the parameter S subjected to t-ratio statistic and the S values between food and money related DDT were subjected to Wilcoxon signed-rank tests for each two-parameter model, so its meaning of different models was verified which could provide an important implication for the food-money related DDT’s description.In study2, a sample of38normal weight adolescents (other possible impulsive disorders has been eliminated) accomplished the food-money related DDT, the eating attitude scale (EAT) and general demographic information assessments. Several measurement indexes about areas and parameters were combined for their measuring significance and applicability analysis. Paired t-tests were used to compare the calculus area (CA) of delay discounting function versus the area under the curve (AUC) for calculation difference between two methods. Paired t-tests were also used to compare areas which come from CA or AUC method for food and money related rewords respectively, and the statistical results’difference between two ways were compared. The difference and the correlation between areas and parameters were analyzesed through the comparison of the fitting curves and the parameters of two groups of discount data with similar CA. Lastely, use these three measurements index compared discount difference of food and money by paired t test.In study3,52healthy-weight (31male,21male) and51overweight/obese (30male,21female) adolescents, sampled from the normal population, were tested on the food-money related Delay Discounting Task (Food-Money DDT), the Door Open Task (DOT), the Barratt Impulsiveness Scale-11(BIS-11),the Eating Attitude Test (EAT),the Eating Disorder Inventory (EDI) and some questionnaires on general demographic information. During the phase of experimental task test, participants also had their height and weight measured for the actual BMI values (BMI=Weight (kg)/height2(m2)). The group delay discounting data of each experimental groups were fitted by the hyperboloid-like model using non-linear regression and the fitted curves were compared. The impulsive variables were compared between groups using multivariable analysis of variance (ANOVA) with weight groups and sociodemographic variables as between-groups factors, and Interaction effect between weight group and sociodemographic variables also was observed. Partial correlation was used to explore the relations among different impulsive variables. Furthermore, the relationships between the BMI and impulsivity dimensions were also established by way of a series of regression analyses.In study4, the measuring tasks and procedure for this study was completely same as study3.84male and59female overweight/obesity adolescents were screened for this study, and some of these were directly come from the study3. One-way ANOVA was used for the difference of AUC and the scores of the eating attitude and the eating disorder scales between male and female groups. Considering the difference of gender differences, the correlation of delay discounting, eating attitude, eating disorder and its subscales was Pearson correlation for analysis. The sample was divided in to High and low groups acording the AUC and the score of the eating attitude, One-way ANOVA was used for eating attitude and disorder difference between high and low discounting degree groups, and was also used for difference analysis of the degree of discounting, eating disorder and each factor of eating disorder among normal, anorexia or bulimia tend group and most likely group.Results:In study1, the comparative results of R2at group and individual level have shown that two-parameter models would fit the discount data more suitability than one-parameter models. The fitting difference between two different two-parameter models was not significant. A portion of cases can’t be fitted effectively by the one-parameter models with exponential decay model nearly about30%and hyperbolic model nearly23%, whereas all cases’discount data could be fitted efficiently by two-parameter models. Nearly about63%of hyperbolic-like model and nearly about58%of the exponential function model whose parameter S were significantly less than1. Used reward as a within-group factor, parameter S was not significant for exponential function model between food and money (Z=-1.756, p=0.079<0.05, but difference significant for hyperbolic-like model (Z=-2.817, p=0.005<0.05, and it can reflect subjects’sensitivity to stimulants for food-money related DDT.In study2, AUC and CA were significant difference on individual specific area about delay discounting(food:t(37)=3.073,p<0.005;￥100: t(37)=4.057, p<0.005), but they are the same when they used for within-group comparison of area between food and money related DDT (CA:t(37)=2.411,p<0.005; AUC:t(37)=2.480, p<0.005),and the two histograms which present the average CA and AUC of all subjects are almost have the same height. So, the area and the parameters which present the delay discounting ability could be associated through the calculus of discount function. When two cases have similar CA, the parameter K can response curve steep degree, and the parameter S can response the curve’s stretching on horizontality with the delay time gone. There is no correlation between CA and LogK (for f or food-DDT: r=0.089, for money-DDT:r=0.055; p>0.05), but moderate negative correlated between CA and LogS (for food-related DDT:r=-0.413, for money-related DDT:r=0.509; p<0.05), and significant negative correlated between LogK and LogS(for food-DDT:r=-0.801, for money-DDT:r=-0.796; p<0.05). The within-group comparison between food and money-related DDT only found significant difference on CA (t=2.411, p<0.05).In study3, the impulse variables were not significant difference between groups which divided by demographic factors, such as age, gender, and family economic income (p>0.80). Simple-effects analyses found that overweight/Obese showed great significant difference with controls on AUC and the N’s of doors opened, they discounted more on food(F=15.37, p<0.01) and money (F=19.28, p<0.01), and opened more doors (F=10.74, p<0.01). No such differences were seen on other behavior impulsivity variables, total BIS and each subscale of it (p>0.05).Interaction analysis between experiment groups and demographic factors found a significant Genderxweight status interaction on AUC’s (for food-related DDT:F(1,101)=10.42, p=0.00; for money-DDT:F(1,101)=9.48, p=0.00) and the N’s of doors opened (F(1,101)=0.433,P=0.05).The significant difference of AUCs between overweight/obese and normal weight individuals was found only for female (for food-related DDT:F=26.38, P<0.01; for money-related DDT:F=44.86, P<0.01) but not for male (p>0.167). Furthermore, the significant difference of the N’s of doors opened was found both in female(F=4.80, P<0.05) and male(F=5.90; P<0.01), and the different degree was bigger in female. There is no correlation among three impulsive tasks (p>0.05). BMI showed a significant negatively correlation with AUCs(r=-0.229and-0.240for food-DDT and money-DDT, respectively), and BMI was uniquely associated with AUC for delay discounting for food (accounting for26.4%of the variance) and for money (accounting for33.4%of the variance).In study4, the correlation between AUC and eating attitudes, eating disorders, and the related all subscales was significant (p>.05) in femal but not in male overweight/obese. In female, AUC and the eating disorder tendency of slimming, gluttony, perfectionism and introspection (person r were-.268,-0.323,-0.297,-0.263respectively). The significant difference of AUC for money-DDT, total EDI and most of its subscales was found among the normal eating group, the bulimia or binge eating tendency group and possible group (p<0.05). Therefore, the overweight/obese female accompanied with bulimia or binge eating was discounted more and displayed more eating problems. High and low discounted groups only show significant difference on bulimia (F=3.752, p<.05) and introspection (F=4.098, p<.05).Conclusion:1. Comparing the one-parameter models, more discounting data would be fitted effectively and the R2was higher by the two-parameter models. That had shown the two-parameter models displayed low requirements and good description for discounting data. It was reasonable to add the parameter S for two two-parameter models. The cases of parameter S which significant less than1were more by hyperbolic-like model fitting, and they could reflect the sensitivity of the subjects on delayed reinforcers. All in all, the hyperbolic-like model could provide best description of food-money retaliated DDT.2. AUC and CA are on the same theoretical basis, they all reflect total discount level in a period of delay time. The parameters and the area were associated by the calculus of discounting function, and represented different aspects of discounting ability. In this way, it can solve describe difficult about same discounted areas but have different curve shapes.3. The results supported the hypothesis that overweight/obese individuals were more impulsive than the normal weight controls on behavior tasks but not on self-report scale. The impulsive difference between experiment groups performance different in different gender. To be specific, first, the overweight/obese subjects showed greater delay discounting than the normal-weight did, but this conclusion can’t be replicated in males; second, the verweight/obese subjects opened more doors than normal-weight did for the N’s of doors opened in both gender, but the difference were more significant in female. Three tasks reflect different impulsiviness dimensions. The relationship between BMI and delay discounting was closer, BMI has good predictive validity on discount level, the greater BMI (that is obesity), and the more tends to choose instant reward.4. There was some gender difference on the relation of delay discounting and eating variables. The relationship between them was more closely in female overweight/obese adolescents but no significant in male overweight/obese adolescents. The impulsive behavior which was reflected by DDT has a close relation to eating behavior and eating psychological characteristics, mainly reflected in the behavior of gluttony and three psychological characteristics aspects of tend to thin body, perfectionism and introspection. High impulsivity overweight/obese female students show serious gluttony symptoms and lower levels of emotional introspection. Moreover, the overweight/obese female students with anorexia or bulimia are more impulsive and even more serious in eating problems than the normal eating ones.更多还原