探索性因子分析及其在应用中存在的主要问题

探索性因子分析的发展非常迅速,已成为教育与社会心理学领域中最常用的统计方法之一。本文全面介绍了探索性因子分析的基本原理,阐述了其发生的机制及基本过程,对其在教育、心理领域应用中存在的问题进行了总结,并针对应用中样本容量和观测变量数目不够、因子求解方法的误用、因子数目的确定标准及因子旋转中存在的问题、因子值缺乏重复验证性、研究结果呈现形式不规范、过于依赖SPSS、缺乏主动性等问题提出了一些相应的建议。     

认知诊断计算机自适应测验中平衡属性收敛的新方法

提出两种认知诊断计算机自适应测验下平衡属性收敛的新方法（MABI、RTA）,模拟研究系统探讨和比较了此二者与已有方法（ABI、IABI和RABI）的表现。结果发现：（1）新方法较不考虑属性收敛的方法有更高的准确率以及更均衡的题目使用率;（2）新方法较ABI和RABI有稍低的准确性,但有更平衡的题目使用率;（3）新方法与IABI的准确性和题目使用率在不同选题策略下各有合优势。总之,两种新方法较好地兼顾测量准确性、题目使用率以及题库曝光情况。     

认知诊断计算机自适应测验中平衡属性收敛的新方法

提出两种认知诊断计算机自适应测验下平衡属性收敛的新方法（MABI、RTA）,模拟研究系统探讨和比较了此二者与已有方法（ABI、IABI和RABI）的表现。结果发现：（1）新方法较不考虑属性收敛的方法有更高的准确率以及更均衡的题目使用率;（2）新方法较ABI和RABI有稍低的准确性,但有更平衡的题目使用率;（3）新方法与IABI的准确性和题目使用率在不同选题策略下各有合优势。总之,两种新方法较好地兼顾测量准确性、题目使用率以及题库曝光情况。     

题目因素分析：基于SEM和基于IRT的两类方法

心理和教育测量一般只能达到顺序量表的水平,其测量数据与被测因子间并非简单线性关系。题目因素分析是用来描述测量题目与因子间非线性关系的统计模型。题目因素分析主要有基于结构方程模型和基于项目反应理论两类方法,两类方法之间存在紧密的联系,甚至可以看作是同一模型的两种表现形式。本文详细阐述了该关系,同时对两类方法在参数估计、模型拟合指标、测量一致性检验和支撑软件等方面的特点进行了分析和比较,以便研究者选择最为适合其研究的方法。     

不同条件下拟合指数的表现及临界值的选择

在本模拟研究中设计了6种样本容量,6种因子载荷,和4种评分等级,并考察了正态和非正态分布两种情况。采用的错误模型为参数误置（真模型中每个因子各由5个题目来测量,错误模型中则是第一个因子由6个题测量,另两个因子各由4个和5个题来测量,即有一个因子载荷被误置）模型。结果发现（1）样本量、载荷量、评分等级数和分布形态都对GOF的取值确有影响。其中分布形态的影响最大。NNFI、IFI在不同条件下的平均值是最稳定的,其次是CFI、RMSEA和SRMR。它们都算是值得推荐的GOF,尤其是NNFI和IFI。（2）在正态分布中,当样本量≥1000时,根据NNFI、IFI、CFI、RMSEA、SRMR对模型是否拟合做出判断时有很低的两类错误率,在样本量<1000时则不理想。在偏态条件下无论选择哪个GOF两类错误率都很高。（3）采用2指数策略在很多情况下也不能显著降低两类错误率。（4）由于在数据分布非正态,或正态但样本量<1000时是难判断模型是否拟合的。因此我们提出了2界值策略。即为每个GOF确定上下两个界值。低于下界值时可判断模型是不正确的,而高于上界值时则可判断模型是正确的。GOF取值处于上下界值之间时难以判断模型是否拟合,只能说越高拟合的可能性越大。这时就要通过跨样本验证和增加样本量来确定模型是否正确     

小学生图形推理策略发展特点的研究

本研究用我们修订的瑞文推理为材料,对145名儿童图形推理策略进行研究。结果发现：小学生在解决图形推理问题时使用六种策略,它们分别是分析策略、不完全分析策略、知觉分析策略、知觉匹配策略、格式塔策略和自主想象策略;不同年龄儿童在解决不同类型题目时的策略使用表现出不同的特点,儿童在解决数量规则题中,知觉分析策略在整个小学儿童阶段占主导地位,而在解决加减规则题中,分析策略占主导地位,随年龄增长而呈上升趋势;小学二年级开始出现图形推理能力发展的飞跃,二年级儿童开始能够同时观察到两种规则,五六年级儿童更能够不受题目形式的影响．而是从本质上把握逻辑规则。     

多维测验项目参数的估计：基于SEM与MIRT方法的比较

作者简要回顾了SEM框架下分类数据因素分析(CCFA)模型和MIRT框架下测验题目和潜在能力的关系模型, 对两种框架下的主要参数估计方法进行了总结。通过模拟研究, 比较了SEM框架下WLSc和WLSMV估计方法与MIRT框架下MLR和MCMC估计方法的差异。研究结果表明：(1) WLSc得到参数估计的偏差最大, 且存在参数收敛的问题; (2)随着样本量增大, 各种项目参数估计的精度均提高, WLSMV方法与MLR方法得到的参数估计精度差异很小, 大多数情况下不比MCMC方法差; (3)除WLSc方法外, 随着每个维度测验题目的增多参数估计的精度逐渐增高; (4)测验维度对区分度参数和难度参数的影响较大, 而测验维度对项目因素载荷和阈值的影响相对较小; (5)项目参数的估计精度受项目测量维度数的影响, 只测量一个维度的项目参数估计精度较高。另外文章还对两种方法在实际应用中应该注意的问题提供了一些建议。     

贝叶斯题组随机效应模型的必要性及影响因素

题组模型可以解决传统IRT模型由于题目间局部独立性假设违背时所导致的参数估计偏差。为探讨题组随机效应模型的适用范围, 采用Monte Carlo模拟研究, 分别使用2-PL贝叶斯题组随机效应模型(BTRM)和2-PL贝叶斯模型(BM)对数据进行拟合, 考虑了题组效应、题组长度、题目数量和局部独立题目比例的影响。结果显示：(1) BTRM不受题组效应和题组长度影响, BM对参数估计的误差随题组效应和题组长度增加而增加。(2) BTRM具有一定的普遍性, 且当题组效应大, 题组长, 题目数量大时使用该模型能减少估计误差, 但是当题目数量较小时, 两个模型得到的能力估计误差都较大。(3)当局部独立题目的比例较大时, 两种模型得到的参数估计差异不大。     

结构方程模型的应用及分析策略

差不多所有心理、教育、社会等概念,均难以直接准确测量,结构方程（SEM,Struc－tulalEquationModelling）提供一个处理测量误差的方法,采用多个指标去反映潜在变量,也令估计整个模型因子间关系,较传统回归方法更为准确合理。本文主要用一系列有关学习动机的虚拟例子,指出每个问题的主要分析策略,以展示SEM在教育及心理学可以应用的研究范畴。文内探讨的方法包括：验证性因素、高阶因子、路径及因果分析、多时段（multiwave）设计、单形模型（SimpleModel）、及多组比较等。     

预测视角下双因子模型与高阶模型的模拟比较

双因子模型和高阶因子模型,作为既有全局因子又有局部因子的两个竞争模型,在研究中得到了广泛应用。本文采用Monte Carlo模拟方法,在模型拟合比较的基础上,比较了效标分别为外显变量和内潜变量时,两个模型在各种负荷水平下预测准确度的差异。结果发现,两种模型在拟合效果方面无显著差异;但在预测效度方面,当效标为显变量时,两个模型的结构系数估计值皆为无偏估计;而效标为潜变量时,高阶因子模型表现优于双因子模型:高阶因子模型的结构系数为无偏估计,双因子模型的结构系数估计值则在50%左右的情况下存在偏差。     

The use of item parcels in structural equation modelling: Non‐normal data and small sample sizes

Maximum likelihood estimation in confirmatory factor analysis requires large sample sizes, normally distributed item responses, and reliable indicators of each latent construct, but these ideals are rarely met. We examine alternative strategies for dealing with non‐normal data, particularly when the sample size is small. In two simulation studies, we systematically varied: the degree of non‐normality; the sample size from 50 to 1000; the way of indicator formation, comparing items versus parcels; the parcelling strategy, evaluating uniformly positively skews and kurtosis parcels versus those with counterbalancing skews and kurtosis; and the estimation procedure, contrasting maximum likelihood and asymptotically distribution‐free methods. We evaluated the convergence behaviour of solutions, as well as the systematic bias and variability of parameter estimates, and goodness of fit.     

Does practice in category learning increase rule use or exemplar use—or both?

Categorization research has demonstrated the use of both rules and remembered exemplars in classification, although there is disagreement over whether learners shift from one to the other or use both strategies simultaneously. Theoretical arguments can motivate predictions for both rule use and exemplar use increasing with more practice. We describe a single large experiment (n = 190) that manipulated the number of training items (category size), the number of presentations of each training item, and the similarity between the training and the transfer stimuli in order to discover when rules and exemplars are most likely to be used. Results showed that rules and exemplars both influenced classification and that exemplars were used more often with smaller categories, with more training on items, and when test items were similar to training items. There was no consistent evidence of a shift from rule-based to exemplar-based categorization with more learning. Importantly, we found a number of conditions in which rules and exemplars were both used, even within individual participants. We discuss our results in terms of hybrid models of classification.     

Bayesian Estimation for Item Factor Analysis Models with Sparse Categorical Indicators

Psychometric models for item-level data are broadly useful in psychology. A recurring issue for estimating item factor analysis (IFA) models is low-item endorsement (item sparseness), due to limited sample sizes or extreme items such as rare symptoms or behaviors. In this paper, I demonstrate that under conditions characterized by sparseness, currently available estimation methods, including maximum likelihood (ML), are likely to fail to converge or lead to extreme estimates and low empirical power. Bayesian estimation incorporating prior information is a promising alternative to ML estimation for IFA models with item sparseness. In this article, I use a simulation study to demonstrate that Bayesian estimation incorporating general prior information improves parameter estimate stability, overall variability in estimates, and power for IFA models with sparse, categorical indicators. Importantly, the priors proposed here can be generally applied to many research contexts in psychology, and they do not impact results compared to ML when indicators are not sparse. I then apply this method to examine the relationship between suicide ideation and insomnia in a sample of first-year college students. This provides an important alternative for researchers who may need to model items with sparse endorsement.     

The association between children's numerical magnitude processing and mental multi-digit subtraction

Children apply various strategies to mentally solve multi-digit subtraction problems and the efficient use of some of them may depend more or less on numerical magnitude processing. For example, the indirect addition strategy (solving 72–67 as “how much do I have to add up to 67 to get 72?”), which is particularly efficient when the two given numbers are close to each other, requires to determine the proximity of these two numbers, a process that may depend on numerical magnitude processing. In the present study, children completed a numerical magnitude comparison task and a number line estimation task, both in a symbolic and nonsymbolic format, to measure their numerical magnitude processing. We administered a multi-digit subtraction task, in which half of the items were specifically designed to elicit indirect addition. Partial correlational analyses, controlling for intellectual ability and motor speed, revealed significant associations between numerical magnitude processing and mental multi-digit subtraction. Additional analyses indicated that numerical magnitude processing was particularly important for those items for which the use of indirect addition is expected to be most efficient. Although this association was observed for both symbolic and nonsymbolic tasks, the strongest associations were found for the symbolic format, and they seemed to be more prominent on numerical magnitude comparison than on number line estimation.     

Coping with bullying in the workplace: the effect of gender, age and type of bullying

A study of bullying, victimisation and the coping strategies employed to tackle it is presented in the first study of bullying in the workplace conducted in Iceland. Participants were 398 members of a union of store and office workers and members of a national organisation of bank-employees. A factor analysis of bullying items identified two factors: general bullying and work-related bullying. Males score higher on both factors, but when asked directly if they have been bullied or not, no significant gender difference appears. A measure of coping strategies when faced with bullying is presented. A factor analysis and multidimensional scaling of these strategies identified four clusters: assertive response, seek help, avoidance and do nothing. These can be arranged on a passive vs. active dimension, which may reflect the severity or duration of the bullying situation. Multiple regressions indicate that males seek help less and use avoidance less than females, and males are more likely to use assertive strategies, confirming gender stereotypes about what constitutes appropriate behaviour. Increased bullying is also associated with the use of avoidance and passive response (do nothing). These results are mainly in accordance with findings from schools. It is suggested that active coping styles are employed during the initial stages of bullying but that victims resort to more passive coping strategies as the bullying becomes more serious. The need for a longitudinal design is proposed to understand the relationship between duration and severity of the bullying experience and the choice of coping strategies.     

基于GRM模型的BP神经网络参数估计

目前参数估计多采用统计方法,存在耗时长、要求被试样本容量大和项目数多等缺点。本文将BP神经网络和降维法相结合,对GRM的项目参数和考生能力参数进行估计。蒙特卡洛模拟结果显示：（1）不管是人多题少还是题多人少,该网络设计下的参数估计精度都较高;（2）可以应用到多个不同等级评分的参数估计中,甚至是超过15个等级的项目参数,估计精度也较高,这是其他参数估计方法所不可比拟的;（3）运行的时长和统计估计方法相比大大缩减。     

Strategic influences on recollection in the exclusion task: Electrophysiological evidence

One assumption underlying the use of the exclusion task as part of the process dissociation procedure is that studied items are successfully excluded only when they are recollected. The present study employed event-related potentials (ERPs) to demonstrate that successful exclusion does not necessarily require recollection. In two experiments, the study tasks for to-be-excluded items were identical, but the tasks employed withtarget items differed, giving better memory for these items in Experiment 1 than in Experiment 2. Successfully excluded items elicited the ERP signature for recollection—theleft parietal old/new effect—in Experiment 2 only. These findings indicate that the subjects adopted different retrieval strategies in the two experiments. It is suggested that they made more use of source information about to-be-excluded items in the second experiment than in the first.     

Feeling the ease: How the use of oral motor fluency changes in amnesia

The present study examined the evolution observed in amnesic patients’ use of motor fluency when making recognition memory decisions. In this experiment, 9 patients with amnesia and 18 matched controls were presented with two recognition memory tasks composed of 3 types of items: (1) natural words, (2) nonwords difficult to pronounce, and (3) nonwords easy to pronounce, the latter having been shown to be processed in a surprisingly fluent manner as long as participants can articulate them at a subvocal level (i.e., oral motor fluency). Our results provide evidence that the motor-movement manipulation was successful to induce a fluency effect. More specifically, data revealed that both amnesic patients and control participants showed a pattern of response consistent with the use of fluency as a cue to memory for studied items. However, only control participants relied on fluency to increase their rate of “yes” responses for unstudied items. These results suggest that patients with amnesia set a more conservative response criterion before relying on oral motor fluency, showing a pattern consistent with the idea that fluency is only used as a cue to memory when it exceeds a certain threshold. These findings are discussed in terms of adaptative metacognition strategies implemented by amnesic patients to reduce fluency-based memory errors as well as in terms of the variations that seem to occur in these strategies depending on the type of fluency that is experienced.     

问题解决测验中过程数据的特征抽取与能力评估

基于计算机的问题解决测验可以实时记录被试探索环境和解决问题时的详细行动痕迹, 并保存为过程数据。首先介绍了过程数据的分析流程, 然后从问题解决测验入手, 分别对过程数据的特征抽取和能力估计建模两方面的研究进行了梳理和评价。未来研究应注意：提高分析结果的可解释性; 特征提取时纳入更多信息; 实现更复杂问题情景下的能力评估; 注重方法的实用性; 以及融合与借鉴不同领域的分析方法。