Induction of Questionnaire Design and SPSS Statistical Analysis in Graduation Design Project

Graduation design projects every year are inseparable from user research methods, especially the use of market survey method based on questionnaire. Questionnaire surveys are mainly used to measure behavioral data and user attitude data that cannot be obtained from background big data, as well as based on user groups to validate and evaluate issue prioritization.

The basic design method of the questionnaire can be summarized as one formula + two question types + three avoidances + four theories. “One formula” is to lock in the questionnaire outline (first purpose of questionnaire = dimension A + dimension B + dimension C). “Two question types” are yes-no questions and closed-questions. For yes-no questions, avoid using the word “no” and balance positivity and negativity. Closed-questions must follow exhaustiveness and mutual exclusivity, and retain exit items. Avoid extreme options, pay attention to the order of options, and put those that meet social expectations last. “Three avoidances” are to avoid tendentious descriptions, avoid vague wording, and avoid double questions. “Four theories” are that high-frequency behaviors use frequency questions, and low-frequency behaviors use possibility questions; high-frequency behaviors have a small time span, and low-frequency behaviors have a large time span; when the number of a certain behavior exceeds 5 times, it is an estimate instead of counting for the user; when asking about the frequency or number of times, use less fill-in-the-blanks and more distributed options commonly.

The questionnaire outline of “First purpose of questionnaire = dimension A + dimension B + dimension C” is actually just a main structure of the questionnaire, which dismantles the problem in a multi-dimensional manner. A key point in questionnaire design is to break down the specific questions and options from dimensions based on the questionnaire outline. Each question needs to be written with a purpose and analysis method to ensure the rationality of the questionnaire design.

After the first draft of the questionnaire is completed, it needs to be pre-tested, usually with a sample of 20-100 people. Using these sample data also requires testing the reliability and validity of the questionnaire. Specific testing methods can be obtained through literature search. For example, you can refer to the article “调查问卷的信度分析和效度分析“.

Several points need to be emphasized here. The split-half reliability method is not suitable for measuring factual questionnaires. It is often used for reliability analysis of attitude and opinion questionnaires, such as Likert scale questions. Alpha reliability coefficient method is mainly used to evaluate the consistency of continuous variables and ordinal variables, such as Likert scale questions. The retest reliability method tests the same group of subjects twice, and then calculates the correlation coefficient of the scores obtained from the two tests, such as categorical variable questions. General fact-finding questions and dimensionless single questions are not subject to reliability analysis. The statistical analysis of content validity mainly uses the single item and sum correlation analysis method to obtain the evaluation results, that is, calculate the correlation coefficient between each item score and the total item score, and judge whether it is valid based on whether the correlation is significant; if there are negative questions in the scale, items should be reversely processed before calculating the total score. Structural validity is the most commonly used validity index, and the method used to analyze structural validity is factor analysis.

The SPSS statistical analysis process of questionnaire results includes four steps: defining variables, data entry, statistical analysis and result saving. There are as many variables as there are questions in a questionnaire, and the answer to each question is the value of the variable. Variables in multiple-choice questions are often defined using multiple dichotomies. The basic idea is to set each option of the question into a variable, and then split each option into two options, that is, check the item and not check the item. When entering data, one row represents one questionnaire, so if there are several questionnaires, there will be several rows of data. There are two types of statistical analysis: graphing analysis and numerical analysis. In SPSS, except for the survival curve graph used in survival analysis, which is integrated into the Analyze menu, other statistical drawing functions are placed in the Graph menu, and the numerical statistical analysis process is done all in the Analyze menu.

参考译文

毕业设计课题中的问卷设计与SPSS统计分析归纳

关键词：毕业设计，问卷，SPSS

每年的毕业设计课题都离不开用户研究方法，尤其是以问卷为主的市场调查方法的运用，问卷调查主要是用于测量后台大数据获取不到的行为数据和用户态度数据，以及根据用户分群来验证和评估问题的优先级。

问卷基础设计方法可以概况为一个公式+两种题型+三个避免+四个理论。“一个公式”为锁定问卷大纲（问卷第一目的=维度A+维度B+维度C）。“两种题型”为是否题型和封闭式题型，是否题型避免使用“不”字，平衡肯定性和否定性；封闭式题型需遵循穷尽性、互斥性，保留退出项，避免选项极端化，注意选项排序，符合社会期许的放在最后。“三个避免”为避免倾向性描述、避免措辞模糊、避免双重问题。“四个理论”为高频行为用频次提问、低频行为用可能性提问；高频行为时间跨度小、低频行为时间跨度大；当某个行为次数超过5次以上，用户不会进行计数，而是估算；提问频次或次数时，少用填空式，常用分布式选项。

“问卷第一目的=维度A+维度B+维度C”的问卷大纲实际只是一种问卷主体结构，是以多维度方式对问题进行拆解。问卷设计的一个关键点就是依据问卷大纲，从维度拆分出具体问题和选项后，每个问题都需要写上目的和分析方法，确保问卷设计的合理性。

问卷初稿完成后需要进行预先测试，通常选择20-100人的样本。利用这些样本数据还需要对问卷的信度和效度进行检验，关于具体的检验方法可以通过文献检索获取，如下文可以参考“调查问卷的信度分析和效度分析”。

这里需要强调几点，分半信度法不适合测量事实性问卷，常用于态度、意见式问卷的信度分析，如里克特量表题型。Alpha信度系数法主要用于评价连续变量和顺序变量的一致性，如里克特量表题型。重测信度法对同一组被试者先后两次进行测查，然后计算两次测查所得分数的相关系数，如分类变量题型。一般事实性调查题型和无维度的单个题目不做信度分析。内容效度的统计分析主要采用单项与总和相关分析法获得评价结果，即计算每个题项得分与题项总分的相关系数，根据相关是否显著判断是否有效；若量表中有反意题项，应将其逆向处理后再计算总分。结构效度是最常使用的效度指标，结构效度分析所采用的方法是因子分析。 问卷结果的SPSS统计分析过程包含定义变量、数据录入、统计分析和结果保存四个步骤。一份问卷有多少个问题就要有多少个变量与之对应，每一个问题的答案即为变量的取值。多选题的变量常常采用多重二分法定义，其基本思想是把该题每一个选项设置成一个变量，然后将每一个选项拆分为两个选项项，即选中该项和不选中该项。数据录入时，一行代表一份问卷，所以有几份问卷，就要有几行的数据。统计分析有作图分析和数值分析两类，在SPSS中，除了生存分析所用的生存曲线图被整合到Analyze菜单中外，其他的统计绘图功能均放置在Graph菜单中，数值统计分析过程均在Analyze菜单中完成。