复习生成网站REVIEW GENERATOR

5.1Qualitative vs Quantitative 定性 vs 定量

The distinction is based on the kind of information used; both are useful depending on the question. 区别在于所用信息的种类；用哪种取决于研究问题，两种都有用。

Exploratory / descriptive questions lean qualitative; explanatory / predictive questions lean quantitative. 探索性/描述性问题偏定性；解释性/预测性问题偏定量。

5.2Research Design — Definition 研究设计的定义

An activity-based and time-based plan, always based on the research question. 以活动和时间为基础的计划，始终源于研究问题。

• Guides the selection of sources and types of information. 指导信息来源与类型的选择。
• A framework for specifying relationships among the study's variables. 界定研究变量之间关系的框架。
• Outlines procedures for every research activity. 列出每项研究活动的程序。

Design answers: what techniques to gather data, what sampling, how to handle time and cost constraints. 设计要回答：用什么技术收集数据、如何抽样、如何处理时间与成本约束。

5.3Classification by Stage 按阶段分类

5.4Methods of Data Collection & Analysis 数据收集与分析方法

Collection 收集：A. Monitoring 观察/监测（like observational studies 类似观察研究） · B. Interrogation / Communication 询问/沟通 · C. Archival sources 档案资料

Analysis 分析：Statistics 统计 · Verbal argument 言语论证 · Ratio Analysis 比率分析 · Logical Analysis 逻辑分析 · NLP 自然语言处理 · AI 人工智能

5.5Power to Produce Effects 产生效应的能力

5.6Purpose 研究目的

• Descriptive 描述性：answer who, what, when, how much; discovery of associations among variables; cross-tabulations and bivariate correlations. 回答谁、什么、何时、多少；发现变量间的关联（交叉表、双变量相关）。
• Causal 因果性：how one variable affects another; A produces / forces B. 一个变量如何影响另一个；A 产生/迫使 B。
• Predictive 预测性：attempt to estimate what will happen in the future. 尝试估计未来会发生什么。

5.7Topical Scope 主题范围

5.8Time Dimension 时间维度

5.9Exploratory Techniques (4) 探索性技术（四种）

• 1) Qualitative techniques / Experience surveys 定性技术/经验调查：in-depth or elite interviews, observation (inductive — generate new ideas); interview/observe a variety of people, not only key persons. Experience surveys ask: what is being done, what was tried, what changed, barriers, costs, priorities. 深度/精英访谈与观察（归纳式，产生新想法）；访谈、观察各类人，不只关键人物。经验调查问：在做什么、试过什么、变了什么、障碍、成本、优先级。
• 2) Secondary data analysis 二手数据分析：prior studies, documents — essentially a literature review. 既有研究与文档——本质上就是文献综述。
• 3) Focus groups 焦点小组：composition better if homogeneous; if participants are heterogeneous, consider splitting; quality depends crucially on the moderator's ability to trigger discussion. 成员同质更好；异质时考虑分组；质量关键取决于主持人引发讨论的能力。
• 4) Two-stage designs 两阶段设计：stage 1 explore to find a feasible research problem; stage 2 explore to find appropriate designs (or further exploration/explanation). 第一阶段探索可行的研究问题；第二阶段探索合适的研究设计（或进一步探索/解释）。

5.10Causation 因果关系

Essential element: A produces B / A forces B to occur. 核心要素：A 产生 B / A 迫使 B 发生。Three conditions 三个条件：

• Covariation 共变：if A occurs, B also occurs; if A is absent, B is absent; less A = less B. A 出现 B 也出现；A 不在 B 也不在；A 减少 B 也减少。
• Time order 时间顺序：A must occur before B. A 必须发生在 B 之前。
• Exclusiveness 排他性：no other possible causes for B. B 没有其他可能的成因。

5.11Confounding Variable / Confounder 混杂变量

A type of extraneous variable related to both the independent and the dependent variables. 一种与自变量和因变量都相关的外来变量。Two conditions 两个条件：

• 1) Must be correlated with the independent variable (may or may not be causal). 必须与自变量相关（可以是因果关系，也可以不是）。
• 2) Must be causally related to the dependent variable. 必须与因变量有因果关系。

Classic example 经典例子：temperature 气温 sits behind both ice-cream sales and drowning. 气温同时在冰淇淋销量和溺水事件背后。

6.1Unit of Analysis 分析单位

Levels 层级：Countries 国家 · Sectors 行业 · Firms 公司 · Departments 部门 · Teams 团队 · Employees 员工 · Decisions 决策

Depends on 取决于：the research problem (at what level do we look for answers) · at what level we need/measure information · at what level we want to implement the answers. 研究问题（在哪一层级寻找答案）、在哪一层级需要/测量信息、在哪一层级落实答案。

6.2Core Concepts 核心概念

Population 总体 · Population Element 总体元素 · Sample 样本 · Census 普查（special case: the whole population is the sample 特例：整个总体即为样本）

6.3Why Sample Instead of Census? 为什么抽样而不普查

• Lower costs + faster 成本更低、速度更快。
• Greater accuracy 准确性可能更高 — trade-off: asking everybody vs obtaining better/more data from a representative sub-group. 权衡：问所有人，还是从有代表性的子集获得更好/更多的数据。
• Availability of population elements 总体元素是否可获得。

6.4What Is a Good Sample? 什么是好样本

• Accuracy 准确性 = absence of systematic variance (bias). 没有系统性偏差。
• Precise estimate 精确估计 = sampling error should be random error; a sample will by chance differ from the population (no way to avoid it), and this decreases with sample size. 抽样误差应只是随机误差；样本因随机机会与总体不同在所难免，且随样本量增大而减小。

6.5Six Steps in the Sampling Process 抽样流程六步

• 1) What is the relevant population? 相关的总体是什么？
• 2) What are the parameters of interest? 关注的参数是什么？
• 3) What is the sampling frame? 抽样框是什么？
• 4) What is the type of sample? 样本类型是什么？
• 5) What size sample is needed? 需要多大的样本？
• 6) How much will it cost? 成本多少？

6.6Key Concepts for Probability Sampling & Sample Size 概率抽样与样本量的关键概念

Standard Error 标准误 · Confidence Interval 置信区间 · Central Limit Theorem 中心极限定理

6.7What Determines Sample Size? 什么决定样本量

• Greater dispersion / variance within population → larger sample needed. 总体内离散/方差越大 → 样本要越大。
• Greater desired precision → larger sample. 想要的精确度越高 → 样本越大。
• Greater number of subgroups of interest → larger sample (each subgroup must meet minimum size). 关注的子群越多 → 样本越大（每个子群都要达到最小量）。
• If calculated sample size exceeds 5% of the population, sample size may be reduced without sacrificing precision. 若算出的样本量超过总体的 5%，可适当减小而不损失精确度。

6.8Probability Sampling 概率抽样

Every element has a known/even chance of selection → statistical inference OK. 每个元素被抽中的机会已知/均等 → 可以做统计推断。

• Simple Random Sampling 简单随机抽样：random selection, e.g. from an email list. 随机抽取，如从邮件名单中随机选。
• Stratified Sampling 分层抽样：divide frame into strata (age, income…) and sample within each stratum. 把抽样框分成层（年龄、收入等），在每一层内抽样。
• Cluster Sampling 整群抽样：randomly select groups (schools, neighbourhoods…), then sample within clusters. 先随机选中群（学校、社区等），再在群内抽样。
• Multistage Sampling 多阶段抽样：combines stages, e.g. regions → cities → households. 组合多个阶段，如地区→城市→住户。

6.9Non-probability Sampling 非概率抽样

Selection based on availability, judgment, referral, or quotas → no statistical inference. 依据可得性、判断、转介或配额选择 → 不能做统计推断。

• Convenience 便利抽样：whoever happens to visit your survey page. 恰好访问你问卷页面的人。
• Purposive / Judgmental 目的/判断抽样：researcher selects based on expertise. 研究者凭专业判断挑选。
• Snowball 滚雪球抽样：participants refer others; useful for hard-to-reach groups. 参与者转介他人；适合难以触达的群体。
• Quota 配额抽样：fill predefined categories until quotas are met. 按预设类别填满配额。

6.10Internet Sampling Caveats 网络抽样的注意事项

• Populations are unclear 总体不明确。
• Is the Internet sample representative? 网络样本有代表性吗？
• Control over who enters the sample? 谁进入样本，不受控制。

7.1Who vs How 问谁 vs 怎么问

Chapter 6 answers “who to ask” 问谁; Chapter 7 answers “how to ask / collect” 怎么问、怎么收集. 第 6 章管挑人，第 7 章管收集方式。

7.2Factors Influencing Participant Motivation 影响参与者动机的因素

7.3Web Surveys — Attractions 网络问卷的吸引力

• Short turnaround of results; results tallied as respondents complete surveys. 结果周转快；填答完成即自动统计。
• Ability to use visual stimuli 可以使用视觉刺激材料。
• Ability to do numerous surveys over time (repeatable tracking) 可随时间重复做多次（可追踪）。
• Attract participants who wouldn't join other projects, including international respondents 能吸引不愿参加其他研究的人，包括国际受访者。
• Respondents feel anonymous 受访者感到匿名。
• Shortened turnaround from survey draft to execution 从问卷草稿到执行的周期更短。
• Experiences unavailable by other means 能提供其他方式无法实现的体验。

7.4Web Surveys — Drawbacks 网络问卷的缺点

• Recruiting the right sample is costly and time-consuming 招募合适的样本既贵又耗时。
• Unlike phone and mail, no sample frames / lists exist 与电话、邮寄不同，没有现成的抽样框/名单。
• Converting surveys to the Web can be expensive 把问卷搬上网成本可能很高。
• Takes technical as well as research skill to field a Web survey 需要技术与研究双重能力。
• Browser compatibility — technology isn't perfect across browsers 浏览器兼容性——技术在不同浏览器上并不完美。