Replication Data for Reconceptualising dimensions of political competition in Europe: A demand side approach

Included are: 1. The raw data (before cleaning and preprocessing) can be found in the files ending "Raw3". The codebooks for each of these data files end in "codebook". This will enable the user to identify the statements that are associated with the items EU1 … 7, Eco1 … 7, Cul1 … 7, AD1 and AD2 that are used in the manuscript.// 2. The R codes ending cleaning_plus.R are used to a) clean the datasets according to the procedure outlined in the online Appendix and b) remove entries with missing values for any of the variables that are used in the calibration process to produce balanced datasets (age, education, gender, political interest). Because of step b), the new datasets generated will be smaller than the clean datasets listed in Table 1 of the Appendix.// 3. For the balancing and calibrating (pre-processing), we use a) the datasets for each country generated by 2 above (the files that are followed by the suffix "_clean"), b) the file drop.py, which is the code (in python) for the balancing algorithm that is based on the principle of raking (see the online Appendix), c) the R files that are used to generate the new calibrated datasets that will be used in the Mokken Scale analysis in 5 below (followed by the suffix "balCode"), and d) a set of files ending in the suffix "estimates" that contain the joint distributions derived from the ESS data (i) for age, below versus above the median age and (ii) for education, degree versus no degree, as well as the marginal distributions for gender and political interest. The median ages of the voting population derived from ESS are as follows: Austria: 50 Bulgaria: 52 Croatia: 52 Cyprus: 47 Czech Republic 50 Denmark: 50 England: 53 Estonia: 50 Finland: 54 France: 55 Germany: 53 Greece: 50 Hungary: 49 Ireland: 50 Italy: 50 Lithuania: 53 Poland: 50 Portugal: 52 Romania: 46 Slovakia: 52 Slovenia: 52 Spain: 50// 4. A set of data files with the suffix myBal, which contain the new calibrated datasets that will be used in the Mokken Scale analysis in 5 (below).// 5. A set of R codes for each country, beginning with the prefix "RCodes" that are used to generate the findings on dimensionality that are presented in the manuscript.

本数据集包含以下内容：1. 未经过清洗与预处理的原始数据存放在后缀为"Raw3"的文件中；各数据文件对应的编码手册（codebook）后缀为"codebook"，可帮助研究者识别论文中使用的EU1至EU7、Eco1至Eco7、Cul1至Cul7、AD1及AD2条目对应的表述。2. 后缀为"cleaning_plus.R"的R代码用于完成两项工作：a) 依据在线附录中载明的流程对数据集进行清洗；b) 剔除校准流程中所需变量（年龄、受教育程度、性别、政治兴趣）存在缺失值的样本。经步骤b处理后生成的新数据集规模将小于附录表1中列出的清洗后数据集。3. 数据集的加权校准（预处理）流程使用以下资源：a) 由步骤2生成的各国数据集（后缀为"_clean"的文件）；b) "drop.py"文件，该Python代码实现了基于raking（见在线附录）原理的加权校准算法；c) 用于生成后续第5步莫肯量表分析（Mokken Scale Analysis）所需校准后数据集的R脚本（后缀为"balCode"）；d) 后缀为"estimates"的文件集合，其中包含基于欧洲社会调查（ESS）数据得到的联合分布：(i) 按年龄中位数划分的低于/高于中位数年龄分组，(ii) 按受教育程度划分的有学位/无学位分组，同时包含性别与政治兴趣的边缘分布。由ESS数据得到的各国投票人口年龄中位数如下：奥地利：50；保加利亚：52；克罗地亚：52；塞浦路斯：47；捷克共和国：50；丹麦：50；英格兰：53；爱沙尼亚：50；芬兰：54；法国：55；德国：53；希腊：50；匈牙利：49；爱尔兰：50；意大利：50；立陶宛：53；波兰：50；葡萄牙：52；罗马尼亚：46；斯洛伐克：52；斯洛文尼亚：52；西班牙：50。4. 后缀为"myBal"的数据集文件集合，其中包含第5步莫肯量表分析所需的新校准后数据集。5. 各国专属的R代码集合，前缀为"RCodes"，用于生成论文中呈现的维度分析结果。