Course Title: PY2501 Research Methods and Data Analysis
Focus: Questionnaire Design: Items and Scales
Instructor: Prof Andrew Schofield
Content Based On: Slides by Prof Adrian Burgess
Week 3: Theory of Measurement
Week 8: From Items to Scales
Week 9: Evaluation of Scale Quality and Interpretation
Avoid Redundancy: Do not create new scales if reliable ones already exist.
Essentials: Good scales must be both reliable and valid.
Challenges: Achieving these attributes in scale creation can be difficult.
Main Topics:
Transition from items to scales
Development of items for measurement
Scaling and response methods
Psychological traits are often multidimensional, shown through:
Physiological (e.g., GSR, ECG responses)
Behavioral (flight/fight/freeze response)
Cognitive/Affective (subjective feelings)
Subjective phenomena require direct inquiry (asking individuals).
From Psychological Theory: Start with hypotheses around characteristics (e.g., extroversion).
Systematic Search for Items: Utilize resources like dictionaries and expert consultations.
A well-defined theory leads to specific questions:
Example (Extroversion):
"Do you enjoy going to parties? (Y/N)"
Include reverse-scaling items (e.g., "I prefer quiet activities.")
Items must reflect all dimensions of a construct (e.g., fear of spiders). Example questions:
Behavioral: "I feel the urge to flee when I see a spider."
Physiological: "Seeing a spider makes my heart race."
Cognitive/Affective: "I often worry about encountering spiders."
Simplicity: Questions must only ask one thing (avoid compound questions).
Readability: Aim for clarity understandable to an average 8-year-old.
Cultural Appropriateness: Ensure language is suitable across different cultures.
Scaling Definition: Transform qualitative attributes into quantitative variables.
Types of Scaling:
Comparative Scaling (e.g., paired comparison)
Direct Estimation (e.g., rating scales)
Formulate criteria for evaluating attributes (e.g., lecturer quality).
Paired Comparison Method: Estimates item differences using judges’ preferences.
Select relevant items to compare.
Engage judges for preferences on pairs.
Conversion of preferences to statistical measures like z-scores.
Eliciting direct responses quantified (e.g., assessing fear levels).
Response Formats: Various scales such as Likert, visual analogue scales, etc.
The number of response steps must balance information richness and statistical reliability.
Labeling of Categories: Important for clarity; endpoints must always be labeled.
Statistical Operations: Assumption allows for parametric testing (means, standard deviations).
Limitations: Be cautious about claims regarding values (e.g., IQ scales).
Measurement Goals: Use multiple items to reduce error and increase reliability (errors are random).
Analyze correlations between items to ensure internal consistency (e.g., Cronbach's Alpha).
Item-Total Correlation: Correlation between single items and total scores.
Split-Half Reliability: Comparing results from two halves of the items.
Cronbach’s Alpha: Statistical measure for overall consistency across items.
Recognize that some psychological scales measure multiple underlying variables (e.g., personality traits).
Use Factor Analysis to identify latent factors and correlations among items.