二参数逻辑斯蒂模型项目参数的估计精度

项目参数的估计精度对于测验的编制尤其是题库的建立十分重要。目前,国内外对项目参数估计精度的研究,大部分是基于在已知项目参数真值的情况下,运用各种参数估计方法产生新的估计值,再和真值进行偏度(BIAS)和均方根差(RMSE)的比较,从而说明该种估计方法的有效性。但是这种方法不能提供不同的参数真值之间的估计误差的变化规律。为了弥补这一缺陷,本文尝试从项目参数估计信息函数的角度出发研究项目参数的估计精度问题。本研究以二参数Logistic模型作为研究对象,首先定义了项目参数的估计信息函数,然后基于完全随机实验设计,通过模拟研究的方法探索影响项目参数的估计精度的因素,实验共设计了(2×3×2)种情形。研究结果表明：(1)项目参数(a,b)的估计精度均随着被试样本量的增大而提高;(2)被试的能力分布对难度参数的估计精度影响较大,对区分度参数的估计精度影响相对较小;(3)难度参数和区分度参数的估计精度都分别受到参数a和参数b的共同作用。     

Knowledge of Results for Motor Learning: Relationship Between Error Estimation and Knowledge of Results Frequency

The authors of the present study investigated the apparent contradiction between early and more recent views of knowledge of results (KR), the idea that how one is engaged before receiving KR may not be independent of how one uses that KR. In a 2 × 2 factorial design, participants (N = 64) practiced a simple force-production task and (a) were required, or not required, to estimate error about their previous response and (b) were provided KR either after every response (100%) or after every 5th response (20%) during acquisition. A no-KR retention test revealed an interaction between acquisition error estimation and KR frequencies. The group that received 100% KR and was required to error estimate during acquisition performed the best during retention. The 2 groups that received 20% KR performed less well. Finally, the group that received 100% KR and was not required to error estimate during acquisition performed the poorest during retention. One general interpretation of that pattern of results is that motor learning is an increasing function of the degree to which participants use KR to test response hypotheses (J. A. Adams, 1971; R. A. Schmidt, 1975). Practicing simple responses coupled with error estimation may embody response hypotheses that can be tested with KR, thus benefiting motor learning most under a 100% KR condition. Practicing simple responses without error estimation is less likely to embody response hypothesis, however, which may increase the probability that participants will use KR to guide upcoming responses, thus attenuating motor learning under a 100% KR condition. The authors conclude, therefore, that how one is engaged before receiving KR may not be independent of how one uses KR.     

A Bayesian Approach to More Stable Estimates of Group-Level Effects in Contextual Studies

Multilevel analyses are often used to estimate the effects of group-level constructs. However, when using aggregated individual data (e.g., student ratings) to assess a group-level construct (e.g., classroom climate), the observed group mean might not provide a reliable measure of the unobserved latent group mean. In the present article, we propose a Bayesian approach that can be used to estimate a multilevel latent covariate model, which corrects for the unreliable assessment of the latent group mean when estimating the group-level effect. A simulation study was conducted to evaluate the choice of different priors for the group-level variance of the predictor variable and to compare the Bayesian approach with the maximum likelihood approach implemented in the software Mplus. Results showed that, under problematic conditions (i.e., small number of groups, predictor variable with a small ICC), the Bayesian approach produced more accurate estimates of the group-level effect than the maximum likelihood approach did.     

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。     

‘I feel better when…’: An analysis of the memory-experience gap for peoples’ estimates of the relationship between health behaviours and experiences

Objective: People often overestimate how strongly behaviours and experiences are related. This memory-experience gap might have important implications for health care settings, which often require people to estimate associations, such as “my mood is better when I exercise”. This study examines how subjective correlation estimates between health behaviours and experiences relate to calculated correlations from online reports and whether subjective estimates are associated with engagement in actual health behaviour.

Design: Seven-month online study on physical activity, sleep, affect and stress, with 61 online assessments.

Main Outcome Measures: University students (N = 168) retrospectively estimated correlations between physical activity, sleep, positive affect and stress over the seven-month study period.

Results: Correlations between experiences and behaviours (online data) were small (r = ?.12–.14), estimated correlations moderate (r = ?.35–.24). Correspondence between calculated and estimated correlations was low. Importantly, estimated correlations of physical activity with stress, positive affect and sleep were associated with actual engagement in physical activity.

Conclusion: Estimation accuracy of relations between health behaviours and experiences is low. However, association estimates could be an important predictor of actual health behaviours. This study identifies and quantifies estimation inaccuracies in health behaviours and points towards potential systematic biases in health settings, which might seriously impair intervention efficacy.     

自尊水平、事件效价与时距对过去事件时距估计的影响

在西方文化背景下提出的时间自我评价理论认为,人们为了维护当前积极自尊,主观上会远离过去失败而亲近过去成功。本研究采用线段评估法,选取311名中国大学生被试,通过两个实验考察自尊水平、事件效价与时距对过去事件时距估计的影响,以检验该理论的文化普适性。结果表明:对过去事件的时距估计,自尊水平、事件效价及其交互作用在同一时距里均不存在显著差异;时距的主效应及其与事件效价的交互作用均显著,在负效价水平上存在显著时距差异,在正效价水平上无显著时距差异。研究结果不支持基于西方文化背景下的时间自我评价理论。     

改进的MCMC算法及其在估计IRT模型参数中的应用

本文首先简要的阐述了MCMC算法的思想及在IRT参数估计中的操作过程;其次,针对该算法存在的一些问题,提出相应的改进建议;然后,分别运用传统的和改进型的MCMC算法进行模拟数据分析和比较,结果显示新的方法表现更好;最后总结新方法的优点所在,并指出下一步的研究方向。     

短时距估计中的心理影响因素

使用单任务研究程序,预期式研究范式,以产生时距作为反应指标,通过两个实验考察短时距估计中的间断位置、间断时距和提示线索效应,探讨产生时距与等待时距的关系问题,揭示注意和间断期望对短时距估计的影响。研究结果表明, 间断实验中表现出极其显著的提示线索效应,间断位置（等待时距）效应显著,间断时距效应及间断位置与间断时距的交互效应不显著。由此得出：(1)注意是影响被试时间判断的主要因素,被试的产生时距随着等待时距的增加而延长。(2)被试对间断的期望显著地影响其时间估计行为,无间断实验条件下,间断期望有损于时间估计。     

Benchmark‐based strategies in whole number line estimation

In this study, we used verbal protocols to identify whether adults spontaneously apply quartile‐based strategies or whether they need additional external support to use these strategies when solving a 0–1,000 number line estimation (NLE) task. Participants were assigned to one of three conditions based on the number of external benchmarks provided on the number line. In the bounded condition only the origin and endpoint were indicated, the mid‐point condition included an additional external benchmark at 50%, and in the quartile condition three additional external benchmarks at 25%, 50%, and 75% were specified. Firstly, participants in the bounded condition reported to spontaneously apply quartile‐based strategies to calibrate their estimates. Moreover, participants frequently relied on the external benchmarks for creating internal benchmarks at the mid‐point, quartiles, and even octiles of the number line. Secondly, overall estimation accuracy improved as the number of external benchmarks increased, and target numbers close to external benchmarks were estimated more accurately and with less variability. Thirdly, the use of a larger variety in benchmark‐based strategies was positively related to NLE accuracy. In summary, this study provides evidence that the NLE task induces more sophisticated strategy use in participants than initially anticipated.