PSY2013 Lecture 04 Notes
PSY2013 Research Methods in Psychology
Ch05 Introduction to Experimental Research
Ch06 Methodological Control in Experimental Research
What is an Experiment? 什么是实验?
- Experiment 实验
- 研究调查至少一个自变量 (IV) X 对至少一个因变量 (DV) Y 的影响,同时保持所有其他因素不变。
- Independent Variable (IV) 自变量 (IV): The manipulated variable that is believed to be the cause.
- 被认为是原因的操作变量。
- Dependent Variable (DV) 因变量 (DV): The measured variable that is the outcome.
- 结果的测量变量。
- Extraneous Variables 外在变量: Factors held constant.
- 保持不变的因素。
Expanded Notes 扩展笔记:
- Woodworth’s ‘Columbia Bible’ (1938): First distinguished between independent and dependent variables in experiments and showed preference for experimental over correlational research.
- Woodworth 的《哥伦比亚圣经》(1938) 首次在实验中区分了自变量和因变量,并显示了对实验研究的偏好。
Correlation and Causation 相关性与因果关系
Correlation 相关性
- Does X occur when Y also occurs? (X <-> Y)
- X 出现时 Y 是否也出现?
- 例子:冰淇淋消费与溺水死亡;巧克力消费与诺贝尔奖。
Causation 因果关系
- Does X cause Y? (X -> Y)
- X 是否导致 Y?
- 例子:锻炼是否导致健康?吸烟是否导致肺癌?
Expanded Notes 扩展笔记:
- Criterions for Causation 因果关系的标准
- Association 关联: Two variables are observed to vary together.
- 两个变量被观察到一起变化。
- Time Order 时间顺序: One variable comes first in time before the other variable.
- 一个变量在时间上先于另一个变量。
- Nonspuriousness 非虚假性: Other possible explanations can be ruled out.
- 其他可能的解释可以排除。
- Mechanism 机制: The process that creates a causal connection between two variables.
- 创造两个变量之间因果关系的过程。
- Association 关联: Two variables are observed to vary together.
The Logic of Experimental Research 实验研究的逻辑
- John Stuart Mill’s Inductive Logic 约翰·斯图尔特·密尔的归纳逻辑
- Method of Agreement 一致法
- If X, then Y (sufficiency -> X is sufficient for Y).
- 如果有 X,则有 Y(充分条件 -> X 足以导致 Y)。
- Method of Difference 差异法
- If not X, then not Y (necessity -> X is necessary for Y).
- 如果没有 X,则没有 Y(必要条件 -> X 是 Y 所必需的)。
- Together, these methods enable us to conclude, with a high degree of confidence (but not absolute certainty), that X causes Y.
- 结合这些方法,我们可以高度确信(但不是绝对确定)X 导致 Y。
- Method of Agreement 一致法
Essential Features of Experimental Research 实验研究的基本特征
Establishing Independent Variables (IVs) 确立自变量 (IV)
Types of IVs 自变量的类型
- Situational 情境性: Different conditions or contexts.
- 不同的条件或情境。
- Task 任务性: Different tasks participants perform.
- 参与者执行的不同任务。
- Instructional 指导性: Different instructions given to participants.
- 给参与者的不同指示。
- Situational 情境性: Different conditions or contexts.
Experimental Group 实验组
- Given treatment.
- 接受处理。
Control Group 控制组
- Treatment withheld.
- 不接受处理。
Expanded Notes 扩展笔记:
- Manipulate Situational IV 操作情境性自变量
- Example: Personal space invasions in the lavatory by Middlemis (1976).
- 例子:Middlemis (1976) 在洗手间的个人空间侵犯实验。
- Manipulate Task IV 操作任务性自变量
- Example: Bushman and Anderson (2009) study on video games.
- 例子:Bushman 和 Anderson (2009) 关于电子游戏的研究。
- Manipulate Instructional IV 操作指导性自变量
- Example: Priming study by Dijksterhuis and Knippenberg (1998).
- 例子:Dijksterhuis 和 Knippenberg (1998) 的启动研究。
Controlling Extraneous Variables (EVs) 控制外在变量 (EV)
- Confounds 混淆变量
- Any uncontrolled extraneous variable.
- 任何未控制的外在变量。
- Results could be due to IV or due to confound.
- 结果可能是由于自变量或混淆变量。
Expanded Notes 扩展笔记:
- Example 例子
- Distributed practice study with different study schedules.
- 不同学习时间表的分布式练习研究。
Measuring Dependent Variables (DVs) 测量因变量 (DVs)
- Dependent Variables 因变量
- Any behaviors measured in an experiment.
- 在实验中测量的任何行为。
- 回顾测量尺度(名义、顺序、间隔、比率)。
Expanded Notes 扩展笔记:
- Problems 问题
- Ceiling Effects 天花板效应: Task is too easy, all scores very high.
- 任务太容易,所有分数都很高。
- Floor Effects 地板效应: Task is too difficult, all scores very low.
- 任务太难,所有分数都很低。
- Ceiling Effects 天花板效应: Task is too easy, all scores very high.
Subject Variables 受试者变量
Definition 定义
- Subject Variables 受试者变量
- Already-existing attributes of subjects in a study.
- 研究中受试者的既有属性。
Expanded Notes 扩展笔记:
- Examples 例子
- Gender, age, religion, SES, personality characteristic.
- 性别、年龄、宗教、社会经济地位、性格特征。
Drawing Conclusions with Subject Variables 使用受试者变量得出结论
- With a Manipulated IV 使用操作自变量
- Assuming no confounds -> IV causes DV.
- 假设没有混淆变量 -> 自变量导致因变量。
- With a Subject Variable 使用受试者变量
- Groups may differ in several ways -> IV cannot be said to cause DV.
- 各组可能在多方面有所不同 -> 不能说自变量导致因变量。
Expanded Notes 扩展笔记:
- Example 例子
- Self-esteem study with manipulated IV (feedback on test performance) and subject variable (low vs. high self-esteem).
- 自尊研究,使用操作自变量(测试表现反馈)和受试者变量(低 vs. 高自尊)。
The Validity of Experimental Research 实验研究的效度
Types of Validity 效度类型
Statistical Conclusion Validity 统计结论效度
- The extent a researcher uses statistics properly and draws the appropriate conclusions from the statistical analysis.
- 研究人员适当使用统计方法并从统计分析中得出适当结论的程度。
Construct Validity 构念效度
- The extent to which the abstract construct can be inferred from the operational definition.
- 从操作定义推断抽象构念的程度。
External Validity 外部效度
- The extent to which research findings generalize to contexts other than those of the experiment.
- 研究发现推广到实验以外环境的程度。
Internal Validity 内部效度
- The degree to which an experiment is methodologically sound and confound-free.
- 实验在方法上是否健全和没有混淆变量的程度。
Expanded Notes 扩展笔记:
- Threats to Internal Validity 内部效度的威胁
- History 历史效应: Results from a historical or external factor affecting the experimental group.
- 由于影响实验组的历史或外部因素导致的结果。
- Maturation 成熟效应: Participants naturally change over time, independent of their participation in the study.
- 参与者随着时间自然变化,与他们参与研究无关。
- Regression to the Mean 均值回归: 极端分数在随后的测试中往往会趋向于平均值。
- Testing and Instrumentation 测试和仪器效应: 随着时间的推移,参与者或测量工具的变化。
- Subject Selection 受试者选择: Groups are not equivalent.
- 各组不等价。
- Attrition 退出效应: Reduction in participant numbers due to dropouts.
- 由于退出导致的参与者人数减少。
- History 历史效应: Results from a historical or external factor affecting the experimental group.
Methodological Control 方法控制
Between-Subjects and Within-Subjects Design 被试间设计和被试内设计
- Between-Subjects Design 被试间设计
- Each subject experiences only one level of the variable.
- 每个受试者仅经历一个变量水平。
- Solved by random assignment or matching.
- 通过随机分配或匹配解决。
Random Assignment 随机分配
- Definition 定义
- A procedure in which each participant has an equal chance of being assigned to any group in an experiment.
- 一种程序,使每个参与者有平等机会被分配到实验中的任何组。
- Ensures that groups are equivalent at the start of the experiment.
- 确保各组在实验开始时是等价的。
Expanded Notes 扩展笔记:
- Example 例子
- In a study on the effects of a new drug, participants are randomly assigned to either the treatment group or the control group.
- 在一项关于新药效果的研究中,参与者被随机分配到治疗组或对照组。
Matching 配对
- Definition 定义
- A procedure to ensure that groups are equivalent by matching participants on specific characteristics before assigning them to groups.
- 一种通过在分组之前根据特定特征匹配参与者以确保各组等价的程序。
- Used when random assignment is not feasible or when there are known confounding variables.
- 在随机分配不可行或存在已知混淆变量时使用。
Expanded Notes 扩展笔记:
Example 例子
- In a study on the effects of teaching methods on student performance, participants are matched on GPA before being assigned to different teaching methods.
- 在一项关于教学方法对学生表现影响的研究中,参与者在分配到不同的教学方法之前根据 GPA 进行匹配。
Within-Subjects Design 被试内设计
- Each subject experiences all levels of the variable.
- 每个受试者经历所有变量水平。
- Solved by counterbalancing.
- 通过平衡解决。
Counterbalancing 平衡设计
- Definition 定义
- A method to control for order effects in a within-subjects design by varying the order of conditions for participants.
- 一种在被试内设计中控制顺序效应的方法,通过改变参与者的条件顺序。
- Types 类型
- Complete Counterbalancing 完全平衡
- All possible orders of conditions are used equally.
- 所有可能的条件顺序都同等使用。
- Example: For 3 conditions (A, B, C), there are 6 possible orders (ABC, ACB, BAC, BCA, CAB, CBA).
- 例子:对于3个条件(A, B, C),有6种可能的顺序(ABC, ACB, BAC, BCA, CAB, CBA)。
- Partial Counterbalancing 部分平衡
- Only a subset of all possible orders is used.
- 仅使用所有可能顺序的一个子集。
- Often achieved through a Latin Square design.
- 通常通过拉丁方设计实现。
- Complete Counterbalancing 完全平衡
Expanded Notes 扩展笔记:
- Latin Square Design 拉丁方设计
- A special case of partial counterbalancing where each condition appears in each position and precedes and follows each other condition exactly once.
- 部分平衡的特殊情况,其中每个条件出现在每个位置,并且每个条件恰好一次出现在其他条件之前和之后。
- A special case of partial counterbalancing where each condition appears in each position and precedes and follows each other condition exactly once.
Example 例子:
- Example of Counterbalancing 平衡示例
- Testing the effect of different study techniques (reading, summarizing, testing) on memory retention.
- 测试不同学习技术(阅读,总结,测试)对记忆保持的影响。
- Complete counterbalancing: Participants experience all 6 orders of the 3 conditions.
- 完全平衡:参与者经历所有3个条件的6种顺序。
- Partial counterbalancing: Using a Latin Square, participants experience a subset of all possible orders.
- 部分平衡:使用拉丁方,参与者经历所有可能顺序的一个子集。
Controlling for Bias 控制偏见
Experimenter Bias 实验者偏见
- Expectations can influence subject behavior.
- 期望可能影响受试者行为。
- Solution: Automatize routine, standardize procedures, use double-blind procedures.
- 解决方案:自动化常规,标准化程序,使用双盲程序。
Participant Bias 参与者偏见
- Participants change behavior because they are being observed.
- 参与者因为被观察而改变行为。
- Solution: Effective deception, use of manipulation checks, field research.
- 解决方案:有效的欺骗,使用操控检查,现场研究。
Expanded Notes 扩展笔记:
- Example 例子
- Hawthorne Effect: Participants change behavior because they are aware of being observed.
- 霍桑效应:参与者因为知道被观察而改变行为。
Summary 总结
- Experimental research involves manipulating independent variables to observe their effect on dependent variables while controlling extraneous variables.
- 实验研究涉及操控自变量以观察其对因变量的影响,同时控制外在变量。
- Methodological controls include random assignment, matching, counterbalancing, and controlling for biases.
- 方法控制包括随机分配、匹配、平衡和控制偏见。
- Ensuring internal validity is crucial for drawing accurate conclusions from an experiment.
- 确保内部效度对于从实验中得出准确结论至关重要。
PSY2013 Lecture 04 Notes