1/16
Looks like no tags are added yet.
Name | Mastery | Learn | Test | Matching | Spaced | Call with Kai | Chat |
|---|
No analytics yet
Send a link to your students to track their progress
Operationalizing Variables
Variables are anything that can change or vary (e.g., age, gender, aggression, anxiety).
What you’re measuring
Defining them - conceptual (what it means) vs. operational (how it’s measured)
Type of measurement
Construct
Variables that cannot be directly observed; traits, emotions, attitudes
Type of Measurement
Self-report: Interviews or questionnaires, people report their beliefs, behaviour, history, etc. (has social desirability bias)
Video-recordings: hint for stress cues like nail biting, fidgeting, or expressions
Behavioural: observing behaviours, either lab induced or naturally occurring
Physiological: assessment of bodily states, brain imaging, or heart rate monitors
How do you choose?
Previous research, methodological (new tech), feasibility (resource limitation)
Reliability
Consistency
True score - the "real" score on the variable (a person truly has an anxiety level of 50)
Obtained score - the score the measure gives (they score 55 because they were especially stressed that day)
Measurement error - difference between true score and obtained score
We want measurement error as small as possible by using reliable, valid measures.
To assess the reliability of report measures:
Test-retest - If I give the same measurement twice with time in between, will I get a similar result?
Parallel-forms - If there are different forms of the measure, do they all measure the same thing? Scores on two different version of test = same score = high parallel
Internal consistency - Do all the questions on a scale measure the same thing?
Split-half: Compare scores from the first half of the questionnaire with the second half.
Cronbach's alpha (α): Measures how well all the questions are related to each other.
Higher α = better internal consistency
To assess the reliability of observational measures:
Interrater reliability
How consistent the results are with two/more researchers observing the participant
Validity
Accuracy
Face Validity - does the measure look like it’s measuring what it’s meant for
Content Validity - measure all the important components/parts of the construct
Criterion Validity - Convergent: does it correlate with similar measures. Predictive: does it predict future expected outcomes
Discriminant/divergent Validity - does not correlate with unrelated measures.
Example: A depression questionnaire should:
Correlate with other depression tests (good convergent validity).
Not strongly correlate with math ability because they measure different things.
Types of Data
Categorical - each value represents a discrete category where order does not matter (tiger = 1, lion = 2…)
Numerical - each value represents either a real number (age) or a place on a scale/continuum, where order does matter
Categorical Graphs
Pie chart (for simple data = 1), bar graph compares a series of categories as bars (for complex data)
Numerical Graphs
Histogram shows the distribution (bell curve), visual sense of the data; the mean, range, skew, possible outliers
Discrete - finite number of values (dice), binned
Continuous - infinite number of values (height), logged
Scatterplot
Used for two continuous variables, shows relationship
Line Graph
Time Series (changes over time)
Y-Axis
The dependent measure or the variable you're most interested in
Should have a reasonable range
Too broad, you can't see the difference (flat line)
Too limited, large top and bottoms/skyrocketing
X-Axis
Consistent range/scale
You can exaggerate small changes by zooming in
Truncating an Axis
Restricting range to maximize differences (exaggerate differences)
Example:
Scores are 90 and 95, but the graph starts at 89, making the difference look huge.
Expanding an Axis
Using too broad a range to minimize differences (hide differences)
Example:
Scores are 60 and 90, but the graph goes to 1,000, making the difference look tiny.
Ignoring Conventions
Values should go from small to large
Example:
Years are shown out of order, like 2024, 2022, 2023.
Comparing non-equivalent data
Two different Y-axis on same graph
Example:
Comparing temperature and sales using two different y-axes