Week 03: Quantitative Analysis I & II

Causality, Research Design, Survey Methods, and Measurement

SSPS4403 Research Methods

The University of Sydney

Semester 2, 2026

Welcome

Acknowledgement of Country

I would like to acknowledge the Traditional Owners of Australia and recognise their continuing connection to land, water and culture. The University of Sydney is located on the land of the Gadigal people of the Eora Nation. I pay my respects to their Elders, past and present.

Required Reading

Textbook: Kellstedt, Whitten, and Tuch (2022)

This week’s chapters:

Chapter 3: Evaluating Causal Relationships
Chapter 4: Research Design
Chapter 5: Survey Research
Chapter 6: Measuring Concepts of Interest

Today’s Journey

Chapters 3-6: Building Research Foundations

Understanding causality and the four causal hurdles
Research design strategies
Survey research methods
Measurement concepts
Introduction to JASP

Learning Objectives

By the end of today, you will be able to:

Apply the four causal hurdles to evaluate research claims
Distinguish between different research designs
Understand survey methodology fundamentals
Evaluate measurement validity and reliability
Explore data structure in JASP

Causality

Why Causality Matters

Modern sociology fundamentally revolves around establishing whether there are causal relationships between important concepts.

— Kellstedt, Whitten, and Tuch (2022, chap. 3)

Remember

Finding a relationship ≠ Finding a causal relationship

Probabilistic vs. Deterministic

In social science:

Relationships are probabilistic, not deterministic
We can’t predict individual behavior with certainty
Example: Wealth → Tax preferences
- Not: “All wealthy people want lower taxes”
- But: “Wealthy people are more likely to prefer lower taxes”

The Four Causal Hurdles

To claim X causes Y, we must cross four hurdles:

Is there a credible causal mechanism connecting X to Y?
Can we eliminate reverse causation (Y causing X)?
Is there covariation between X and Y?
Have we controlled for confounders (Z variables)?

Hurdle 1: Causal Mechanism

Question: How and why would X cause Y?

Think Through the Process

What is the logical connection?
What are the intermediate steps?
Is the mechanism plausible?

Example: School choice → Test scores

Mechanism: Private schools → Smaller classes → More attention → Better learning → Higher scores

Hurdle 2: Reverse Causation

Question: Could Y cause X instead?

Sometimes easy to rule out:

Gender → Abortion attitudes ✓
- Attitudes can’t cause gender

Sometimes difficult:

Intergroup contact ⟷ Racial tolerance
- Contact may increase tolerance
- OR tolerant people may seek contact
- OR both!

Hurdle 3: Covariation

Question: Do X and Y vary together?

Generally necessary for causation
“More X → More Y” or “More X → Less Y”
BUT: Correlation ≠ Causation!

Important Caveat

No bivariate relationship? Don’t give up yet! A confounding variable Z might be suppressing the relationship.

Hurdle 4: Confounding Variables

Question: Have we controlled for other causes of Y?

The Problem

Most outcomes have multiple causes
If Z causes both X and Y…
The X-Y relationship might be spurious

Example: School Choice

X: School type
Y: Test scores
Z: Parental involvement
- Affects school choice
- Affects test scores

The Causal Hurdles Scorecard

Track answers with: [hurdle1 hurdle2 hurdle3 hurdle4]

y = yes, n = no, m = maybe
[y y y y] = Strong causal claim ✓
[n ? ? ?] = Stop and reformulate
[y n y y] = Reasonable evidence (reverse causation unclear)
[y y y n] = Suspicious! Need to control for confounders

In-class Task

Causal Claims Exercise

In groups, evaluate these claims using the scorecard:

“School choice programs improve student test scores”
“Red wine consumption reduces heart disease”
“Social media use causes depression”

For each claim:

Assess all four hurdles [y/n/m y/n/m y/n/m y/n/m]
Identify the strongest and weakest hurdles
What additional evidence would strengthen the claim?

The four hurdles:

Is there a credible causal mechanism connecting X to Y?
Can we eliminate reverse causation (Y causing X)?
Is there covariation between X and Y?
Have we controlled for confounders (Z variables)?

Research Design

Why Research Design Matters

Good research design helps us:

Cross the four causal hurdles more effectively
Control for confounders (hurdle 4)
Rule out reverse causation (hurdle 2)
Make stronger inferences about causality

Important

Bad design → Bad statistics (no matter how sophisticated!)

Types of Research Designs

Experimental

Random assignment
Researcher controls treatment
Gold standard for causality

Observational

No random assignment
Observe natural variation
More common in social science

Experimental Designs

Key features:

Random assignment to treatment/control groups
Randomization ensures groups are equivalent
Rules out confounders (on average)
Can establish causation with confidence

Example: Testing teaching methods

Randomly assign students to new vs. old method
Compare outcomes
Differences → Caused by method

Observational Designs

Challenges:

Groups differ in many ways
Cannot randomly assign many social factors
- Gender, race, education history, neighborhood, etc.
Must control for confounders statistically
Causation is harder to establish

Example: Effect of poverty on crime

Can’t randomly assign poverty!
Must account for other differences between poor/non-poor neighborhoods

Quasi-Experimental Designs

Middle ground:

Not truly experimental (no random assignment)
BUT: Leverage natural experiments
Policy changes, cutoff dates, geographic boundaries

Example: Voting law changes

Some states change laws, others don’t
Compare turnout before/after
Compare across states

Design → Causal Hurdles

Design	Mechanism	Reverse	Covariation	Confounders
Experimental	Test	✓✓✓	Test	✓✓✓
Quasi-exp.	Test	✓✓	Test	✓✓
Observational	Test	✓	Test	✓

Better design → More hurdles cleared

Group Activity

Design Matching

In groups of 3-4:

For each research scenario, identify:

The most appropriate research design
Why it’s suitable
Which causal hurdles it helps address
Remaining challenges

Scenarios on handout

Survey Research

Why Surveys?

Surveys are a primary tool for:

Measuring attitudes and opinions
Understanding behaviors and experiences
Collecting data from large populations
Enabling generalization

But: Survey design matters!

Sampling: The Big Challenge

The Goal:

Learn about a population (all Australian adults)
By studying a sample (1,000 survey respondents)

Key Question

Is your sample representative of the population?

Types of Sampling

Random Sampling

Every population member has known probability of selection
Allows generalization
Examples: Simple random, stratified, cluster

Non-Random Sampling

Convenience, snowball, quota
Cannot generalize with confidence
May introduce bias

Survey Design Considerations

Question wording
- Clear, unambiguous language
- Avoid leading questions
- Avoid double-barreled questions
Response options
- Exhaustive and mutually exclusive
- Appropriate scale
Question order
- Priming effects matter

Common Survey Challenges

Non-response bias: Who doesn’t respond?
Social desirability bias: Respondents give “acceptable” answers
Recall problems: Memory limitations
Mode effects: Phone vs. online vs. in-person

Tip

Always consider: What biases might affect these survey results?

Measurement

From Concepts to Data

The measurement process:

Concept (abstract idea)

↓

Operationalisation (how to measure)

↓

Variable (actual data)

Example: Measuring “Racism”

Concept: Racism (abstract)
Operationalisation choices:
- Attitudes toward racial groups?
- Support for discriminatory policies?
- Implicit association test?
- Behavioural measures?
Variable: Survey responses, test scores, etc.

Each choice has implications!

Key Measurement Concepts

Two fundamental properties of good measurement:

Validity - Accuracy
Reliability - Consistency

Validity: Measuring What You Intend

Central question: Does this measure actually capture the concept I care about?

The Core Issue

A measure can be reliable but not valid!

Example: Measuring “intelligence” by shoe size - Very reliable (shoe size doesn’t change much) - Not valid (shoe size isn’t intelligence)

Types of Validity

Face Validity

Definition: Does it look like it measures what it claims?
Example (Valid): Asking “How happy are you?” to measure happiness
Example (Invalid): Asking “What’s your favorite color?” to measure happiness
Limitation: Subjective judgment, not rigorous

Content Validity

Definition: Does it cover all aspects of the concept?
Example: Testing math skills
- ✓ Valid: Questions on algebra, geometry, statistics
- ✗ Invalid: Only algebra questions
Method: Expert review of content coverage

Types of Validity (continued)

Construct Validity

Definition: Does it relate to other variables as theory predicts?
Example: Depression measure should:
- Correlate with anxiety (related concept)
- Predict absenteeism (behavioral outcome)
- Differ between clinical and non-clinical groups
Gold standard for validity

Criterion Validity

Definition: Does it predict relevant outcomes?
Example: SAT scores
- Should predict college GPA
- Concurrent: Correlates with current grades
- Predictive: Forecasts future performance

Validity Example: Measuring “Political Knowledge”

Concept: How much do people know about politics?

Low Face Validity: “Do you follow the news?” (could say yes but not retain info)
Better Content Validity: Multiple questions covering:
- Institutions (Who is the Prime Minister?)
- Current events (What party controls Parliament?)
- Policy (What does Medicare cover?)
Construct Validity Check:
- Should correlate with education level ✓
- Should predict voter turnout ✓
- Should be higher among political science majors ✓

Reliability: Measuring Consistently

Central question: Would we get the same result if we measured again?

Key Principle

Without reliability, we can’t have validity!

If a thermometer gives random readings, it can’t accurately measure temperature.

Types of Reliability

Test-Retest Reliability

Definition: Do you get the same score if you measure the same person twice?

Example: Personality test
- Take test on Monday → Score 75
- Take same test on Friday → Score 74
- High correlation = good reliability
Challenge: Some concepts genuinely change (mood, opinions)
Application: Use for stable traits (personality, intelligence)

Types of Reliability (continued)

Inter-Rater Reliability

Definition: Do different observers/coders give the same score?

Example: Coding tweets as “positive” or “negative”
- Coder A: 120 positive, 80 negative
- Coder B: 118 positive, 82 negative
- Agreement = good reliability
Measurement: Cohen’s kappa, percent agreement
Critical for: Content analysis, behavioral observation, essay grading

Types of Reliability (continued)

Internal Consistency

Definition: Do multiple items measuring the same concept give similar results?

Example: Depression scale with 10 questions
- If truly measuring depression, all items should correlate
- Person scoring high on “I feel sad” should also score high on “I feel hopeless”
Measurement: Cronbach’s alpha (α > 0.70 is good)
Application: Multi-item scales and surveys

Reliability Example: Measuring “Life Satisfaction”

Single Item: “How satisfied are you with your life?” (1-10 scale)

Test-retest: Ask same person 2 weeks apart
- Good reliability: Scores within 1 point
- Poor reliability: Monday=8, Friday=3

Multiple Items: Life Satisfaction Scale

“In most ways my life is close to my ideal”
“The conditions of my life are excellent”
“I am satisfied with my life”
“So far I have gotten the important things I want in life”
“If I could live my life over, I would change almost nothing”

All items should correlate (internal consistency)

The Validity-Reliability Relationship

Four possible scenarios:

	Reliable	Unreliable
Valid	✓✓ IDEAL Accurate & consistent	✗ Problematic Can’t be valid without reliability
Invalid	✗ Precise but wrong (like a biased scale)	✗✗ WORST Random noise

Tip

Think of a dartboard: - Valid & Reliable: Darts clustered around bullseye - Reliable but Invalid: Darts clustered, but in wrong spot - Unreliable: Darts scattered everywhere

Practical Implications

When developing measures, ask:

Face validity: Does this seem reasonable?
Content validity: Am I covering all aspects of the concept?
Construct validity: Does it relate to other variables as expected?
Test-retest: Are scores stable over time (if concept is stable)?
Inter-rater: Do different coders agree (if coding required)?
Internal consistency: Do multiple items correlate (if using scale)?

Good measurement is the foundation of good research!

Variable Types

Three measurement levels:

Categorical

Qualitatively different
No inherent order
Example: Religion

Ordinal

Can rank order
Unequal intervals
Example: Satisfaction (high/med/low)

Continuous

Equal unit differences
Mathematical operations meaningful
Example: Age, income

Variable Types Matter!

Why care about measurement level?

Determines appropriate statistics
Affects interpretation
Informs analysis choices

In JASP

You’ll need to specify: Nominal, Ordinal, or Scale

Getting Started with JASP

JASP Interface Overview

What You See When You Start JASP

Left Side

Top Menu Bar
- File, Edit, Data tabs
- Analysis buttons (Descriptives, T-Tests, Regression, etc.)
Data View
- Spreadsheet with rows (cases) and columns (variables)
- Each row = one participant/observation
- Each column = one variable

Right Side

Results Panel
- Analysis output appears here
- Updates in real-time
- Can copy/export results
Analysis Options Panel (center-left when analysis selected)
- Choose variables
- Select options
- Customize analysis

Loading Data into JASP

Two Main Ways to Get Data

Method 1: Open CSV Files (Most Common)

Click File → Open
Select Computer → Browse
Navigate to your .csv file (you can also open Excel files)
Click Open

CSV File Format: - Plain text file - Comma-separated values - Can be created in Excel, Google Sheets, etc. - First row = variable names

Method 2: Open .jasp Files

Contains both data AND previous analyses
Click File → Open → Select .jasp file
(This is also the format to save your JASP workspace)

First Dataset: titanic.csv

Understanding Variable Types

Three Variable Types in JASP

JASP uses symbols in column headers to show variable type:

Type	Description	Example
Scale	Continuous numerical data	Age, Height, Income
Ordinal	Ordered categories	Education level (1=Low, 2=Medium, 3=High)
Nominal	Unordered categories	Gender, Color, Country

IMPORTANT

JASP guesses variable types when loading data
It may guess wrong!
You must check and fix variable types
Wrong variable type → Wrong analysis options → Wrong results!

Changing Variable Types

Two Methods:

Method 1: Click the Icon

Click the symbol in the column header
Select the correct type from dropdown
Done!

Method 2: Use the Variable Editor

Double-click the column header
Change measurement level
Can also add labels here

Important Tips:

IDs should be Nominal (even if numbers)
Likert scales can be Ordinal or Scale (your choice)
Binary variables (0/1) should be Nominal
Age is Scale (continuous numbers)

Hands-On: Loading the Titanic Dataset

Your Task:

Download the file titanic.csv from Canvas
Open JASP on your computer
Load the data:
- File → Open → Computer → Browse
- Find titanic.csv and open it
Examine the data:
- How many rows (passengers)?
- How many columns (variables)?
- What variables do you see?

Expected: 891 rows, 12 columns

Hands-On: Identifying & Fixing Variable Types

The Titanic Dataset Variables:

Look at your data and identify which type each variable should be:

Variable	Current Type?	Should Be?
PassengerId	?	?
Survived	?	?
Pclass	?	?
Name	?	?
Sex	?	?
Age	?	?
Fare	?	?

Your Task:

Check what type JASP assigned to each variable
Decide what type each should be
Fix any incorrect types by clicking the icon

Solution: Correct Variable Types for Titanic

Variable	Should Be	Why?
PassengerId	Nominal	ID number (not meaningful numerically)
Survived	Nominal	Binary: 0 = No, 1 = Yes
Pclass	Nominal or Ordinal	Ticket class: 1, 2, 3 (categories)
Name	Nominal	Text label (categorical)
Sex	Nominal	male or female (categorical)
Age	Scale	Continuous number (0-80 years)
Fare	Scale	Continuous number (price paid)

Key Decisions:

Survived: Could be ordinal, but typically treated as nominal for chi-square tests
Pclass: Could be ordinal (ordered categories) or nominal—both work!

Adding Value Labels in JASP

Making Your Data More Readable

Why Add Labels?

Numeric codes (0, 1) are hard to interpret
Labels make output easier to read
Essential for clear reporting and presentations

Example: Survived Variable

Current: 0 and 1 (confusing!)
Better: “Did not survive” and “Survived” (clear!)

When to Use Labels

Binary variables (Yes/No, Male/Female)
Categorical codes (1=Low, 2=Medium, 3=High)
Any nominal variable with numeric codes

Hands-On: Adding Labels to Survived

Make the Survived Variable More Interpretable

Step-by-Step Instructions:

Open the Variable Editor (Edit Data):
- Double-click on the Survived column header
- OR right-click the column → Edit column
Add Value Labels:
- In the variable editor window, find the Labels section
- Click + Add label
- For value 0, enter label: “Did not survive”
- Click + Add label again
- For value 1, enter label: “Survived”

Apply and Check:
- Click outside the editor or press Enter
- Run a frequency table: Descriptives → Descriptive Statistics
- Move Survived to Variables
- Check Frequency tables
Observe the Difference:
- Your frequency table now shows meaningful labels!
- Instead of “0” and “1”, you see “Did not survive” and “Survived”

Descriptive Statistics

What Are Descriptive Statistics?

Summarizing Your Data

For Continuous Variables (Scale)

Central tendency: Where is the “middle”? - Mean, Median, Mode

Spread: How spread out are the values? - Standard deviation, Variance, Range - Minimum, Maximum, Quartiles

For Categorical Variables (Nominal/Ordinal)

Frequencies: How many in each category? - Counts and percentages

Why Use Them? - Understand your data before analysis - Spot errors and outliers - Describe your sample in reports

Measures of Central Tendency

Finding the “Middle” of Your Data

Mean (Average)

Sum of all values ÷ number of values
Most commonly used
Sensitive to outliers
Example: Ages = 20, 25, 30, 35, 100 → Mean = 42

Median

Middle value when data is sorted
Robust to outliers
Example: Ages = 20, 25, 30, 35, 100 → Median = 30

Mode

Most frequently occurring value
Can be used for any variable type
Example: Ages = 20, 25, 25, 25, 30 → Mode = 25

Which to use?

Normally distributed data → Mean
Skewed data or outliers → Median
Categorical data → Mode

Measures of Spread

How Variable Is Your Data?

1. Standard Deviation (SD)

Average distance from the mean
Most commonly reported
Same units as your data
Larger SD = more spread out

2. Variance

Squared standard deviation
Less intuitive but used in calculations

3. Range

Maximum - Minimum
Shows full spread

4. Interquartile Range (IQR)

Range of middle 50% of data
75th percentile - 25th percentile
Robust to outliers

Computing Descriptives in JASP

Step-by-Step Guide

Steps

Open the Analysis Menu:
- Click Descriptives button (top menu bar)
- Select Descriptive Statistics
Select Variables:
- Drag variables from left box to Variables box
- OR click variable and use arrow button

Choose Statistics

Choose Statistics:
- Under Statistics, check boxes for what you want:
- ☑ Mean
- ☑ Median
- ☑ Std. Deviation
- ☑ Minimum / Maximum
- etc.
View Results:
- Results appear immediately in right panel
- Update in real-time as you change options

Hands-On: Descriptive Statistics for Age

Your Task:

Calculate descriptive statistics for the Age variable:

Steps

Go to: Descriptives → Descriptive Statistics
Move Age to the Variables box
Under Statistics, select:
- ☑ Mean
- ☑ Median
- ☑ Std. Deviation
- ☑ Minimum
- ☑ Maximum
- ☑ Range

Questions to Answer

What is the average age of Titanic passengers?
What is the median age?
What is the youngest age? Oldest?
Is the mean or median higher? What does this tell us?

Interpreting Age Statistics

Expected Results (approximately):

Statistic	Value	Interpretation
Mean	~29.7	Average age around 30 years
Median	~28.0	Middle age is 28
SD	~14.5	Typical deviation is ±14-15 years
Min	~0.42	Youngest passenger was an infant
Max	~80	Oldest was 80 years old
Range	~80	Ages span 80 years

Observations:

Mean > Median → Slight positive skew (right tail)
Some older passengers pull the mean up
Large SD indicates substantial age variation
Note: Some ages are missing (Missing = 177)

Frequency Tables for Categorical Data

Counting Categories

For Nominal/Ordinal Variables

Frequency tables show:

Count in each category
Percentage in each category
Cumulative percentages (for ordinal)

How to Create in JASP

Descriptives → Descriptive Statistics
Move categorical variable to Variables box
Under Tables, check:
- ☑ Frequency tables

Result

Table showing counts and percentages
Helps understand distribution of categories

Hands-On: Frequency Tables

Part A: Survival Status

Create a frequency table for Survived
Answer:
- How many passengers survived?
- How many died?
- What percentage survived?

Part B: Sex Distribution

Create a frequency table for Sex
Answer:
- How many males vs females?
- What percentage were male?

Part C: Passenger Class

Create a frequency table for Pclass
Answer:
- Which class had the most passengers?
- What’s the distribution across classes?

Expected Results: Frequency Tables

Survived

Value	Count	Percent
0 (No)	549	61.6%
1 (Yes)	342	38.4%

Sex

Value	Count	Percent
male	577	64.8%
female	314	35.2%

Pclass

Value	Count	Percent
1 (First)	216	24.2%
2 (Second)	184	20.7%
3 (Third)	491	55.1%

Insights:

Most passengers died (61.6%)
More males than females (about 2:1)
Most passengers were in 3rd class (55%)

Hands-On: Split Frequency Tables

Understanding Survival Patterns Across Groups

The Question: Did survival rates differ by passenger class? By sex?

Part A: Survival by Passenger Class

Go to: Descriptives → Descriptive Statistics
Move Survived to the Variables box
Move Pclass to the Split box
Under Statistics, check: ☑ Frequency tables

What You See:

Separate frequency tables for each class (1st, 2nd, 3rd)
Survival counts and percentages within each class

Observe:

Which class had the highest survival rate?
Which class had the lowest?
What pattern do you notice?

Hands-On: Part B - Survival by Sex

Now You Try:

Using the same approach:

Keep Survived in the Variables box
Replace Pclass with Sex in the Split box
Keep Frequency tables checked

Questions to Answer:

What percentage of males survived?
What percentage of females survived?
Is there a notable difference?
What might explain this pattern?

Note

Discuss: What story do these two analyses tell us about survival on the Titanic?

Drawing Plots with JASP

Visualizing Distributions

Adding Plots to Descriptives

For Scale Variables

Under Basic Plots/Customizable Plots in Descriptive Statistics:

☑ Distribution plots → Histogram
☑ Display density → Smooth curve overlay
☑ Boxplots → Shows median, quartiles, outliers

Benefits

Histograms: See shape of distribution
Density plots: Smooth distribution curve
Boxplots: Quick summary with outliers

For Categorical Variables

Frequency tables are often sufficient
Bar charts can be created elsewhere in JASP

Hands-On: Visualizing Age Distribution

Your Task:

Create visualizations for the Age variable:

Steps

Keep Age in the Variables box
Under Plots, check:
- ☑ Distribution plots
- ☑ Display density
- ☑ Boxplots

Observe and Discuss

What shape is the distribution?
Is it symmetric or skewed?
Where is the peak?
Are there any outliers?
What does the boxplot tell you about the median and quartiles?

Reading a Boxplot

Understanding Box-and-Whisker Plots

Anatomy of a Boxplot

    ●  ← Outlier (beyond 1.5 × IQR)
    |
    ┬  ← Maximum (within 1.5 × IQR)
    |
    ┼──┐
    │  │
    │──│ ← Median (thick line)
    │  │
    └──┼
    |
    ┴  ← Minimum (within 1.5 × IQR)
    |
    ●  ← Outlier

Interpretation

Box: Contains middle 50% of data (25th to 75th percentile)
Whiskers: Extend to most extreme non-outlier values
Dots: Outliers beyond 1.5 × IQR

Quick Assessment: - Symmetric box → Symmetric distribution - Long upper whisker → Right skew - Dots → Potential outliers to investigate

Grouped Descriptive Statistics

Comparing Groups

What if we want to compare subgroups?

Example: Age statistics separately for males and females

How to Do It

Descriptives → Descriptive Statistics
Move continuous variable (Age) to Variables
Move grouping variable (Pclass) to Split box
Select desired statistics

Result: - Separate statistics for each group - Easy comparison across groups - Can also split plots!

This is useful for

Comparing survivors vs non-survivors
Comparing passenger classes
Any categorical breakdown

Hands-On: Comparing Age by Survival Status

Your Task:

Compare age distributions for survivors vs non-survivors:

Steps

Descriptives → Descriptive Statistics
Move Age to Variables
Move Survived to Split
Check: Mean, Median, SD, Min, Max
Under Plots, check Boxplots

Questions

Were survivors younger or older on average?
Is there a difference in median age?
Do the boxplots show different patterns?
What does this suggest?

Discussion: Age by Survival Status

Expected Pattern:

Survived	Mean Age	Median Age	Observations
0 (Died)	~30.6	28	Slightly older
1 (Survived)	~28.3	28	Slightly younger

Insights

Small difference in mean age
Similar median ages
Children may have had better survival rates
But age alone doesn’t tell the whole story!

Next Steps

These are just descriptive statistics
To test if differences are significant, we need:
- Chi-square test (for categorical comparisons)
- t-test (for mean comparisons)
- More complex analyses

This is where hypothesis testing comes in!

Hands-On: Complete Descriptive Analysis

Your Challenge:

Perform a complete descriptive analysis of Fare:

Part 1

Calculate descriptive statistics:
- Central tendency (Mean, Median)
- Spread (SD, Range, IQR)
- Check for missing values
Create visualizations:
- Histogram with density
- Boxplot

Part 2

Compare across passenger classes:
- Split by Pclass
- Compare mean fares
Interpret:
- What is the typical fare?
- How variable are the fares?
- Do fares differ by class?

Creating Computed Variables

Transforming Data in JASP

What are Computed Variables?

New variables created by performing calculations on existing variables
Useful for: transformations, sum scores, z-scores, differences, ratios

Common Uses: - Log transformations → Normalize skewed data - Z-scores → Standardize variables - Sum scores → Combine multiple items - Difference scores → Change from baseline - Means → Average across variables

How to Create

Click the “+” button in the column header row
Name your new variable
Select “Computed type” → “Computed with drag-and-drop”
Enter your formula
Click “Compute column”

Hands-On: Creating Computed Variables

Let’s create two new variables together

First variable: FARE + 1
Then use this new variable to create: log(FARE + 1)

What’s the point of taking the log of FARE + 1?

This is necessary because some passengers paid 0 as FARE. And \log(0) = -\infty.

Of course, -\infty would be impossible to visualize.

Adding 1 to variables that contain zeros is a common, quick-and-dirty trick to still return meaningful visualizations.

Understanding the Log Transformation

What Did We Just Do?

The Formula: log(Fare+1) - Takes the natural logarithm (base e) of each Fare+1 value - Compresses large values more than small values - Result: more symmetric distribution

Example Transformations:

Original Fare	log(Fare+1)	Interpretation
£1	0.00	Log of 1 is 0
£10	2.30
£100	4.61	Less difference at high end
£500	6.21

Notice: The gap between £1 and £100 is huge in original scale, but only 4.61 units in log scale. Meanwhile, £100 to £500 (5x increase) is only 1.6 log units.

Hands-On: Compare Original vs Log-Transformed Fare

Now let’s see the difference:

Steps

Create histograms for both:
- Descriptives → Descriptive Statistics
- Move both FARE AND log(FARE + 1) to Variables
- Under Basic Plots, check:
  - ☑ Distribution plots
  - ☑ Display density
- Under Customizable Plots, check:
  - ☑ Boxplot
Calculate descriptives for both:
- Check: Mean, Median, SD, Min, Max, Skewness

Questions to Answer

Which distribution looks more normal?
How did the skewness change?
Which measure (mean or median) is more similar after transformation?
What happened to the outliers?

Exporting Your Results

Saving Your Work

1. Copy Individual Tables/Plots

Right-click on any result
Select Copy or Copy Special
Paste into Word, PowerPoint, etc.
Tables are APA-formatted!

2. Export Data

File → Export Data
Save as CSV (can open in Excel)
Useful if you’ve computed new variables

3. Save JASP File

File → Save As
Saves data AND all analyses
Can reopen later and continue work
Share with collaborators

Tip

Pro Tip: Save as .jasp file regularly to preserve your work!

Wrap-Up

What We Covered Today

Four causal hurdles for evaluating causal claims
Research designs and their strengths/limitations
Survey research fundamentals and challenges
Measurement concepts: validity, reliability, variable types
JASP basics: interface, loading data, descriptive exploration

Connecting It All

The research process:

Start with causal question (Ch 3)
Choose appropriate research design (Ch 4)
Collect data (Ch 5: surveys, etc.)
Measure concepts carefully (Ch 6)
Explore data before testing (Ch 7, today in JASP)

Looking Ahead: Week 04

Next week: Statistical Analysis

Descriptive statistics in depth
Probability and inference
Hypothesis testing (crossing hurdle 3!)
Bivariate relationships
Introduction to regression

Before Next Week

Familiarise yourself with the JASP interface. Explore datasets on your own!

Questions?

Key takeaways:

Causality requires crossing four hurdles
Research design helps us cross those hurdles
Measurement matters - from concept to variable
JASP is a tool for exploring and testing our data

Preparing for your own research:

Think about your research question - which design would work? What variables would you need?

Thanks!

See you next week!

Resources:

Kellstedt, Whitten, and Tuch (2022, chaps. 3–6)
JASP tutorials: jasp-stats.org/how-to-use-jasp

References

Kellstedt, Paul M., Guy D. Whitten, and Steven A. Tuch. 2022. The Fundamentals of Social Research. Cambridge University Press. https://doi.org/10.1017/9781316415399.