lesson 6 Flashcards
What does Surveys collect?
Individual preferences and evaluations/opinions
advantage and disadvantage of survey?
Advantage: Create your own variables so unique data, fits purpose
(Usually we analyze data created by others)
Challenge: many things can go wrong
One problem of survey?
there is a One-way communication channel.
Cannot ask “how do you mean that?”
what is meat by Planning a survey?
Define clear goal
Plan analysis with survey:
- Target group
- Demographic information
- Expected number of answers number of options per question
Questions Data types
- Binary (yes/no)
- Categorical (unordered/ordered),
- Numerical
- Textual
taking into consideration: People’s time is valuable
how is a Multiple choice – single select?
Select: exactly ONE
Implementation
- Radio buttons (cerchio)
Answer options = choice of partition pizza slicing
Options must be mutually exclusive
Consider “other”
Do not forget parts of the pizza
Do not count parts of the pizza twice
Numeric ranges: approximate logarithmic scale.
Discrete (un)ordered variable
how is a Multiple choice – multiple select?
Select: ZERO to ALL
Software usually allows for this
Option “none” is unnecessary
Implementation
- Check boxes (quadrato)
Answer options multiple yes/no slicing of multiple pizzas
Options NOT mutually exclusive
Set of similar yes/no questions
Do not mix positive and negative options
Series of binary (dummy) variables
how is a Likert scale – ordered slider?
Select: ONE out of several orderd
Numeric slider (Scale 1-5)
Text slider (“not satisfied” … “extremely satisfied”)
Star rating / smiley rating
Special case: thumbs up / down
Options along one dimension, e.g.
Preference (“how important is …”)
Label clearly (“good”/”bad”)
Balanced options: “very good”, “good”, “bad”, “very bad”
Consistency: do not swap “left” and right” between questions
Even number of options = avoid “middle” choice
Discrete ordered
how is Open-end (free text) question?
Flexible(!)
But
- Requires clear question / instructions
- Lower response rate
- All sorts of (potentially similar) answers
- Who will analyze the answers how?
Plan analysis from the beginning
- Manual inspection
- AI-enabled summary
- !!!!Do not ask questions that you cannot analyze.
- Numeric analysis difficult ask qualitative questions
Describe a Survey Design?
- Short Lower probability of incomplete answers.
- Objective No leading questions.
- Simple Use simple, direct language. Few options.
- Thoughtful Have respondents in mind. Explain, but not too much.
- Difficult questions at the end
what can you do Before you hit “send”?
(Once a survey is distributed, it is hard to correct mistakes)
Good: Test
Test You, colleagues, friends.
Collect feedback
“How do you mean this?”
“I find no appropriate choice”
Better: Pilot study with structured interviews
Fixed questions
Open answers
Transcribe, analyze, create answer options
Small N
Distribution
how Avoid sample selection bias?
test with friends
pilot study
what about planning of a survey?
Goal
plananalysis (target)
questions
check avoid bias
distribution