Hedging-Based Scoring Rules for Multiple-Choice Questions

Replication package for "Hedging-Based Scoring Rules for Multiple-Choice Questions", by Jingcheng Fu, Xing Zhang and Songfa Zhong.

Stata data file: FZZ_iq_mcq_sr.dta

Index of the do-files and definition of the necessary globals before running any of the other do files: _master.do

Figures:
Figure 2: Frequency of mixing by condition: sum_iq.do
Figure 3a: Proportion of questions answered with mixing by difficulty level: sum_iq_rule.do
Figure 3b: Proportion of questions answered with mixing by question order: sum_iq_qorder.do
Figure 4: Distribution of IQ score by condition: sum_iq.do
Figure 5: Response time and return on time investment by difficulty level: sum_roti.do
Figure 6: Score and mixing by treatment and risk aversion: sum_risk.do
Figure 7: Score and mixing by treatment and false diversification: sum_fd.do
Figure 8: Spearman correlation between GPA and IQ score: corr_gpa.do

Tables:
Table 1: Determinants of hedging by mixing: reg_iq.do
Table 2: Determinants of standardized IQ score: reg_iq.do
Table 3: Treatment effects conditional on risk attitude: sum_risk.do
Table 4: Treatment effects conditional on false diversification: sum_fd.do
Table 5: Treatment effects conditional on gender: sum_gender.do

Computing Cronbach alpha and confidence intervals: sum_cronbach_alpha.do

Appendix:
Figure A1: Score by difficulty level: sum_iq_rule.do
Figure A2: Score by question order: sum_iq_qorder.do
Figure A3: Response time and ROTI by difficulty level, split by time constraint: sum_roti.do
Figure A4: Summary of raw GPA by university and faculty: corr_gpa.do
Figure A5: GPA-IQ Spearman correlation by randomly drawn question subsets: corr_gpa_subsets.do
Table A1: Interaction effects between time constraint and treatments: reg_iq.do
Table A2: Question-level analysis of mixing: reg_iq_question.do
Table A3: Question-level analysis of IQ scores: reg_iq_question.do
Table A4: Determinants of decision time: reg_iq.do
Table A5: Return on time investment by condition: sum_roti.do
Table A6: Balance checks for post-IQ test characteristics: sum_spillover.do
Table A7: Big Five personality traits summary: sum_big5.do
Table A8: Replication of Table 1 excluding subjects with missing administrative data: reg_iq.do
Table A9: Replication of Table 2 excluding subjects with missing administrative data: reg_iq.do

本数据集为论文《基于套期保值的选择题评分规则》（Hedging-Based Scoring Rules for Multiple-Choice Questions）的复现包，作者为傅景程、张星与钟松发。

Stata数据文件：FZZ_iq_mcq_sr.dta

运行所有其他do文件前的索引脚本与必要全局变量定义文件：_master.do

图表：
图2：按组别划分的混合作答频次：sum_iq.do
图3a：按难度层级划分的混合作答题目占比：sum_iq_rule.do
图3b：按题目顺序划分的混合作答题目占比：sum_iq_qorder.do
图4：按组别划分的IQ分数分布：sum_iq.do
图5：按难度层级划分的作答时长与时间投入回报率：sum_roti.do
图6：按处理组与风险规避程度划分的得分与混合作答情况：sum_risk.do
图7：按处理组与错误多样化程度划分的得分与混合作答情况：sum_fd.do
图8：GPA与IQ分数的斯皮尔曼（Spearman）相关性：corr_gpa.do

表格：
表1：混合作答套期保值行为的影响因素：reg_iq.do
表2：标准化IQ分数的影响因素：reg_iq.do
表3：基于风险态度的处理效应：sum_risk.do
表4：基于错误多样化程度的处理效应：sum_fd.do
表5：基于性别的处理效应：sum_gender.do

用于计算克朗巴哈α系数（Cronbach alpha）及置信区间的脚本：sum_cronbach_alpha.do

附录：
图A1：按难度层级划分的得分情况：sum_iq_rule.do
图A2：按题目顺序划分的得分情况：sum_iq_qorder.do
图A3：按时间约束分组的作答时长与时间投入回报率、按难度层级划分：sum_roti.do
图A4：按院校与院系划分的原始GPA概况：corr_gpa.do
图A5：基于随机抽取题目子集的GPA-IQ斯皮尔曼（Spearman）相关性：corr_gpa_subsets.do
表A1：时间约束与处理组的交互效应：reg_iq.do
表A2：混合作答的题目层级分析：reg_iq_question.do
表A3：IQ分数的题目层级分析：reg_iq_question.do
表A4：决策时长的影响因素：reg_iq.do
表A5：按组别划分的时间投入回报率：sum_roti.do
表A6：后测IQ特征的平衡性检验：sum_spillover.do
表A7：大五人格特质（Big Five personality traits）概况：sum_big5.do
表A8：剔除缺失行政数据被试后的表1复现：reg_iq.do
表A9：剔除缺失行政数据被试后的表2复现：reg_iq.do