Reliability
- Created by: MollyGeehan
- Created on: 06-11-18 10:33
View mindmap
- Reliability
- how consistent the findings from investigation are
- measuring device - most reliable if produces consistent results every time
- if measurement is repeated, then described as reliable
- Ways of testing reliability
- Test-Retest
- same test/ questionnaire to the same person on different occasions
- if it is reliable, then results should be the same each time
- sufficient time between test and retest to ensure pps cannot recall answers
- e.g. questionnaire - two sets of scores correlated to ensure similarity
- positive correlation - reliability of measuring instrument is good
- e.g. questionnaire - two sets of scores correlated to ensure similarity
- sufficient time between test and retest to ensure pps cannot recall answers
- if it is reliable, then results should be the same each time
- same test/ questionnaire to the same person on different occasions
- Inter-rater Reliability
- agreement between two or more observers in behaviour observations
- measured by correlating observations of two or more observers
- (total no of agreements) / (total no of observations) > +.80 --> data have high inter-rater reliability
- measured by correlating observations of two or more observers
- one researcher's interpretation of events may differ from the other's
- subjectivity bias and unreliability in data collection
- Pilot study of observation - to check observers are applying behavioural categories in same way
- both observers must watch same event, record data independently
- agreement between two or more observers in behaviour observations
- Test-Retest
- Improving Reliability
- Questionnaires
- measured using test-retest method
- two sets of data compared should exceed +.08
- one that produces low test-retest reliability requires some items to be rewritten
- e.g. some are complex, interpreted differently by same person, on different occasions
- solution: replace some open questions with closed ones
- Interviews
- use same interviewer each time
- interviewers must be properly trained
- e.g. one is not asking questions that are too ambiguous
- easily avoided in structured interviews
- interviewer's behaviour more controlled by fixed questions
- unstructured interviews less likely to be reliable
- interviewer's behaviour more controlled by fixed questions
- Experiements
- lab experiments described as reliable
- researcher can exert strict control over aspects of procedure
- control is more achievable in lab than in field
- precise replication of particular method than demonstrating reliability of a finding
- one thing affecting reliability - pps tested under slightly different conditions each time
- lab experiments described as reliable
- Observations
- improved by making sure behavioural categories = operationalised, measurable, self-evident
- categories should not overlap & all behaviours should be covered
- categories not operationalised well, different observers make on judgements of what to record
- could end up with differing records
- Questionnaires
- how consistent the findings from investigation are
Comments
No comments have yet been made