Flashcards in Psyc7006 - Video gaming - wk4 Deck (13):
Define: inferential statistics
A test to tell us whether a difference we find between conditions than we expect based on random variation
probability that we would get our results just based on chance if there were no real differences between conditions
The cutoff we use for deciding whether or not an effect is due to chance, 5% (.05) is the convention
Define: Type 1 error
false positive: we find a significant effect even though there isn’t a real difference. We decide an effect is real/significant, even though its actually due to chance (5% of the time if the alpha is .o5)
Define: Type 2 error
false negative: we fail to find a significant effect even though there really is a difference between conditions.
Define: Power (and what determines it)
Sample size. The probability of finding a significant effect in our data if there really is a difference between conditions: 1 minus the probability of type 2 error (e.g., 10% chance of type 2 error = 90%)
Why replicate a study?
To control for type 1 errors
* Maybe the original finding was statistical error (5% chance)
* Analysis errors, selective reporting, and "fishing" can increase the risk of type 1 error (more than 5% chance) i.e., our true alpha may be higher than nominal alpha (.05)
To control for fraud
* Maybe the original researchers lied.
To control for bad research
* Maybe the original study was confounded or poorly designed
To generalise results to a new population or setting
* Maybe the original finding was idiosyncratic to the population or circumstances where it was tested.
To provide a new test of the underlying theory
* Maybe the original finding wasn't a strong enough test, can we design a different experiment to test the same theory?
(*last three goes beyond simple repetition of the experiment)
Reason to avoid a direct replication?
* Results might be difficult to publish - a successful replication is considered boring, a failed replication is assumed to be a mistake. - only successful replications + extensions tend to get published.
* Costs may be prohibitive or resources inaccessible - larges studies may be very expensive to repeat, specialised populations might not be available.
What explains a failure to replicate?
* original finding was type 1 error
* original finding was fraud
* original finding was the result of a confound
* origin finding does not generaliseWhat
* original finding was based on a mistaken theory
* REPLICATION ITSELF MAY NOT HAVE HAD ENOUGH STATISTICAL POWER
Three 'types' of direct replication are? (give definition and examples)
* Direct Replication - we try to recreate the original procedure as closely as possible (provides heterogeneity of irrelevancies)
* Replication and extension - we repeat all or part of the original procedure and add new conditions to extend the findings.
* Conceptual replication - we test the hypothesis of the original study using a new research method. E.G., .....
- 1. Bargh et al., 1996 - young adults who are primed with words about getting old walk more slowly when they leave the lab.
2. Jostmann et al., 2009 - people holding a heavy clipboard make decisions more carefully.
where the linking concept = the notion that unconscious thinking affects our behaviour.
Advantage of conceptual replication = broad generalizability of the phenomenon
Disadvantage of conceptual replication = are the experiments really similar enough to prove or disprove one another? (people have tried to replicate Bargh's experiment unsuccessfully)
Why is there doubt regarding the findings/robustness of a relationship between video games and perceptual cognitive gains?
(hint: statistics, null finding bias, methodology, demand, correlation, lack-of-blindness –hypothesis guessing, overt recruitment, lab using repeat subjects which is not true replication.)
Design a video game replication study that helps control for the methodological flaws that are common in most video game studies
Cross-sectional, subvert recruitment (no mention of video game experience - subject-blindness +researched blindness)...object or spatial attention task (or attentional blink)...video game questionnaire AFTER task. Could include video-game play questionnaires in a large generic survey study and later on recruit those people without telling them the study is about video gaming.
Longitudinal study is more difficult! - control cannot simply be a 'no-game' condition, has to be something equally believable e.g., ...a game like tetris versus a action FPS like unreal tournament.
Recruit participants that have not been involved in other video gaming studies.......