Chapter 7: Causal Analysis §

(F) Day of the week: Tuesday Class: IS303 Created Time: December 15, 2020 2:38 PM Database: Class Notes Database Date: December 15, 2020 2:38 PM Days Till Date: Passed Last Edited Time: June 9, 2021 10:42 AM Type: Lecture

Quick Notes §

3 causal mechanisms

• Time Ordering
• Association
• Non-spurious Association

1. Causal Models §

Two Causal approach

• Nomothethic approach
• Idiographic approach

1.1. Nomothetic §

a general approach that seeks the common factors responsible for events in a research

• More general questions or factors
  • What causes suicide?
• Level: Macro
• Explanations: Probabilistic, a guess, not definite answer
• Data: Quantitative (Deductive)

1.2. Idiographic §

Seeks to find specific factors of the specific event.

• Question: Specific
• Levels: Micro, looking at specific one or few subject rather than a large group
• Explanation: Deterministic, find exact reason and cause
• Data: Qualitative (Inductive, Survey)

A morel limited scope

• What caused Robin to take his own life?

2. Nomothetic Approach §

To study how one variable affect another variable

Must have 3 Requirements

2.1. Temporal Ordering (time ordering) §

Independent variable must be before dependent variable in time

A happens before B, then A is the cause of B

2.2. Association Between Variables §

Two variables picked for research must be related in some way

A $\leftrightarrows$ B

association doesn’t imply cause,

but to cause one another, the two must be associated

2.3. Non-Spurious Association §

We need to pick variables that are directly affecting each other, not through a third party variable.

Spurious: two variable is connected through a third variable. the ‘independent variable isn’t directly leading to the dependent variable

A → B ← C → A

A $\nrightarrow$ C

To control C${}_{(s)}$

• Analyze C’s separation to see if it’s the cause of change in dependent variable
• If A still → B, then A & B survive spurious test

The sea of C${}_{(s)}$: science effort is to control C or eliminate C in the causal model.

3. Causal Analysis in Different Research Designs §

3.1. Experimental Design §

To document effects of independent and dependent variables Research where researchers:

• Experimental Group: exposed to the independent variable
• Control Group: not exposed to the independent variable
• Random Assignment: equal chance to end up in control group or experimental group.

Experimental Design has very strong internal validity

Example: The causal connection between using nicotine patch (Cause) and subsequent changes in smoking behaviors (effects)

• Set up control and experimental group through random assignment
• Expose experimental group to independent variable

3.1.1. Threat to Internal Validity §

What makes a research internally invalid. These can be eliminated by the use of an experiemental design

• History: some large event that could affect subject rather than from the independent variable
• Maturation: when subjects has changed through time influencing dependent variable
• Selection Bias: when Random Assignment isn’t implemented. Subjects are picked in favor for research

3.1.2. The Limits of Experimental Design §

• Uncontrollable Variables: some variables cannot be controlled by the researchers
  • fixed variables: variables attached to subjects since birth (early in life). Cannot be manipulated (gender, race…)
• Non-Ethical Experiment: researchers might create ethical and political issues for conducting the research
• Weak External Validity: findings might not apply in real life society instead of a lab
  • it lacks mundane realism

3.2. Causal Analysis via Non-experimental Survey Research §

causal analysis typically through cross-sectional design: data collect at one point in time

3.2.1. Method/Process §

• time-ordering:

  • Fixed variable as independent variable, which doesn’t change no matter when the data was collected. Since birth
  • Include retrospective questions in questionnairs: questions about the past, unchanging

• non-spurious: Include all possible C variables in the survey

  How would you solve non-spurious by including Cs in survey?

  Look at theory, the connection between variables, which connects to Cs

  • Make C variables the same for all subjects from both control group and experimental group

• association: use statistical test to control C in analysis

  Do literature review

  • to see all Cs
  • to review past independent/dependent variables

---

Survey is used for qualitative questions, but there are weaknesses

• Surveys don’t use experimental design
• Do not manipulate independent variables
• Surveys are not as controlled as an experiment
• If survey delivered to the potential respondents.

3.3. Causalty and Field Research §

Data of causal analysis are words & images, need to respect causal mechanisms

• description/Explorative
• correlational analysis
• exploration or evaluation
• process or change

Analyze to provide ‘thick description’ in order to narrate to specify causal mechanism, contextualize to see which event happened first

Field Research §

To pursue causal analysis

Idiographic Model §

Weakness

• Narrative: to put the story in order, Different individual observing the same event will have different outcome or result
• Constructing an argument doesn’t mean it’s correct