Reliability + Validity Flashcards
(6 cards)
Types of reliability
Internal - measure extent of consistency of experiment. Is all the question measuring for the same thing from different participant.
External reliability - ability to repeat study on same participant at later and still get similar result.
Observer reliability - apply to observation. Is observer consistent and in what they record. Same observation? Classify thing same way?
Way of assessing reliability
Inter observer reliability - 2 or more observer carry out observation same time on same object or same checklist. See correlation between observation with spearman. If positive correlation 0.8 or more = good inter rater reliability
Split half method (counter balancing) - split sets of question in half + participant. One group do half and other do other half. Then vice versa. Use spearman to look for any correlation between results. If similar = high internal reliability
Test - retest - carry out same test on same participant at later data. Then compare using spearman, if similar result = high external reliability
How to improve reliability
Observer: use behavioural checklist. Receive training on how to use coding system/ checklist
Interval - if split half show some result correlate poorly, remove the question
External - use standardised instruction + procedures so it is repeatable
Internal + external relatability
External - extent that findings can be generalised to general public or real world. Not just for participant but to all people. Not artificial environment
Internal - is findings actually caused by independent variables and not extraneous variable
Other validity
Content validity - does the content/ question represent + ask for area of interest. Could ask expert to evaluate + suggest improvement or approve method
Predictive validity - whether findings allow for prediction of behaviour. Can theory found out use to test people for factors + predict behaviour
Face - is question measure what researcher intend to measure. Use intuitive measure to judge
Concurrent - compare test with another test that is established + valid. If produce same result = valid. Could ask participants to carry out both test then compare
Construct - does test measure target construct. Is question relevant to findings want to find. By checking the definition of what findings you are trying to find
Improving validity
Single bind - participant does know what group
-no demand characteristics
Double blind - researcher also not know - avoid researcher bias
Counter balancing (ABBA)
Standardised procedure + instructions
Ensure participant have variety of aspect to be representable
