Lecture 7- Observational Methods 1 Flashcards
Why use observational methods
- Questionnaires of limited applicability
- Apparatus limits generalisability
- Context dependent behaviour where context may be difficult to replicate
What are the steps of an observational stream
- Observe informally
- Choose measures
- Chose recording method
- Collect analyse data
What are the steps of an experimental stream
- Hypothesis
- Predict
- Design
- Experiment
- Analyse
- Interpret
Advice for science
- Ask questions
- Observe informally
- Chose measures
- Don’t code for behaviour that isn’t relevant to your question
- Balance what you want vs what you can do
- When and how do you sample your behaviour
Define the measures with either
- Operational definitions, specify the physical requirements for coding a behaviour
- Ostensive definitions, provide examples through pictures or descriptions
Classify your measures as either
- Events, short duration occurrence
- States, long duration event (sleep)
Types of measures
- Latency, how long the subject takes to respond
- Frequency, countable number
- Rate, frequency per unit time
- Duration, single occurrence time
- Proportion
Scales of measurement
- Non-parametric statistics
- Parametric statistics
Types of non-parametric statistics
- Nominal (categorical)
- Ordinal (ranking)
Types of parametric statistics
- Interval (0 is arbitrary, does not mean not there, temp)
- Ratio-interval (continuous)
Types of sampling rules
- Ad libitum
- Focal sampling
- Scan sampling
- Behaviour sampling
What’s ad libtum
+Preferred method for preliminary observations
+Useful for rare, important events
-Tends to miss rare events of short duration
-Underestimates contribution of smaller subjects
What’s focal sampling
+Specific individual is isolated for observation
-Cab be large if focal subject seeks privacy for certain behaviour
What’s scan sampling
+A number of individuals is sampled (typically an entire group)
- Conspicuous events are overestimated
- Rare events underestimated
What’s behaviour sampling
+Aka all occurrence sampling
-Overestimation of rare events
What are types of recording rules
- Time sampling, can underestimate rare behaviours
- Continuous recording, underestimate long duration behaviours
Coding scheme is a
Measuring instrument
Principles of measurement
- No such thing as perfect measurement
- Measurements are more or less accurate
- Measurements are more or less precise
What is intra observer reliability
The same observer coded the same behavioural record at different times
What’s inter observer reliability
The different observers independently coded the same behaviour
Consensus estimates are based on the assumption
That 2 or more coders can come to exact agreement, typically used on nominal data
Consistency estimates are based on the assumption
That it is unnecessary for 2 or more coders to interpret a scale identically, typically ordinal or continuous data
Types of consensus measurement
- Percent agreement, does not correct for agreement by random chance
- Cohen’s kappa, proportion of agreement after corrections by random chance
Types of consistency measures
- Correlation coefficient, doesn’t take into account variance between coders
- Cronbach’s a, Corrects for variance between coders
How does Cohen’s Kappa work
- Convert frequencies to proportions as the first step
- K= Pa - Pc/ 1 - Pc
- Pa is proportion of agreement
- Pc is expected proportion of agreement
When is kappa value good enough
- 0.41< good agreement
- 0.61< substantial agreement
- 0.81 < near perfect agreement