Test 3 Flashcards
Experiment
allows for inferences of causality *only method that can, mechanisms, and explanation
conditions for causality
- temporal precedence
- consistency, regularity
- contingency - if theres no cause, theres no effect
experimental method
- manipulate the variable (independent)
- control other factors
- measure some result (dependent)
* hard to achieve these goals
Independent variable
- factor of interest
- the potential cause
simplest manipulation
presence and absence
- can compare amount/ levels
- need control of condition of absence cause
ideally manipulation changes only ….
factor of interest
___ should be very similar, which is ___ and impacts ability to _____
control, tough, infer causality
difference between experimental and control conditions =
the factor of interest
IV strengths
control and learn causality. great for things that can be easily changed in the lab
IV weaknesses
1.
2.
3.
4.
5.
6.
- generalizability
- making someone do something
- no longer naturally occurring
- participant reactivity
- is the manipulation realistic/ real world
- does the behaviour change when its being manipulated
extraneous variables
variables not of interest must be controlled. can be done through random assignment.
confound
an EV that covaries with the IV and could provide an alternative explanation. fatal flaw, EV not of interest, only appears when IV is there.
ex. getting a coffee before a test or not getting one (confound). is it getting a coffee or getting a gift?
manipulation check
did you manipulate the IV/cause? strength of association. important for interpreting the results of experiments
between subjects studies
different people in each condition
within subjects studies
same people do the experiment and control
fewer people needed, no possible confound of individual differences, *more powerful, more sensitive/ability to notice differences
between subjects info
1.
2.
3.
- IV id subject variable (age, personality, culture)
- need uninformed people in all conditions (deception)
- participation in one study would affect being apart of the other
between subjects problems
must avoid confound from non-equivalent groups
- use random assignment and larger sample sizes
dependent variable
outcome measures
ceiling effect =
floor effect =
everyone does well
everyone does poorly
*low variability and restriction of range
internal validity =
external validity =
how sound is the study design?
do the results generalize to the real world?
Internal validity
1.
2.
3.
4.
- free of confounds
- influence of task order (fatigue, practice)
- help by counterbalancing - half get control 1st and half get control 2nd
- add a filter task between conditions
IValidity demand characteristics
there may be clues on what is being studied which can affect behaviours
IValidity construct validity
measuring properly
IValidity participants problems
no random assignment (self selection)
selective attrition (participants drop out)
IValidity experimental problems
bias/expectations
help by hiding conditions from experimenter
external validity
1.
2.
3.
4.
5.
- are we learning about the real world?
- populations
- situations/environments
- times, does it happen for different generations?
- generalizing is a limitation, but not a design flaw
other validity concerns
statistical conclusion validity
- appropriate use, adequate sample sizes, selective reporting of analyses
Quasi experiments
control conditions not possible or equivalent - no random assignment
when quasi experiments?
very common in applied context (real world)
interventions/ program evaluation
quasi is not bad why?
necessary when true experiments arent possible
can be more common / useful
a good way to judge if a good true experiment was done - way to critique
one group designs
the most simple
changing the order of events
one group post-test only
collect data after intervention
no comparing because you do not have data from before
no causal inferences and lacks internal validity
one group pre+post-test
DV is tested twice, before and after intervention and compare these scores.
Quasi experiments internal validity threats
1. alternative explanation
2. alternative explanation
3.
4.
- effects of history = some real world event occurs to affect DV, after pre-test before post
- maturation = natural changes that can occur on DV (age and development)
- testing = pre test influences post test (fatigue, practice, participant reactivity, hawthorne
- regression toward the mean = for both high and low scores, they tend toward the average. select a group that has an extreme value and keep testing.
Can control help for quasi?
would help rule out history, maturation, and testing but only if control condition also goes through these things. control also needs extreme scores for regression.
single case quasi experiment
only 1 participants - useful for yourself
measure baseline, then do intervention, then measure measure behaviour
research on ___ does not ____ to ____
groups, generalize, individuals
reversal design
baseline A, treatment B, baseline A
see if you can go back to baseline, casual principle
end on __ if the ___ ____
B, treatment works
developmental designs & types
examining change across ages - naturally-occuring
cross-sectional and longitudinal
cross sectional
1. like….
2. strengths
3.weaknesses
measure interest in people of different ages
1. between-subjects
2. relatively easy, inexpensive
3. differences may not be in age but due to other things (cohorts)
longitudinal design
1. like….
2. strengths
3. weaknesses
same people followed over time
1. same-subjects
2. no confounds with other cohorts
3. difficult, expensive, attrition
- retrospective method
- strengths
- weaknesses
- ask people to recall earlier experiences
- easy, inexpensive
- misremembering or not accurate
Ethics =
research is =
morality, integrity
complex and unpredictable, cannot create a set of rules but follow principles
concern for welfare
minimize harm, maximize benefit. ignorance can result in harm. stay well-trained
respect for persons
1.
2.
3.
- protect autonomy
- free of coercion
- informed consent.
promote justice
1.
2.
3.
- equal access to benefits and burdens
- participant selection justified
- those who contribute must benefit (Jane & Finch)
- Institutional review board
- exempt from review
- expedited from review
- full review
- need to submit proposal to them before research
- naturalistic observation in public and archival
- minimal risk research
- everything else
Minimal risk
stress from study is equal to everyday stress. benefits must outweigh the harm. all research is a burden
Rules for conducting experiments
1.
2. the burden of this
3.
- need voluntary, informed, competent consent
- informing may alter behaviour - deception but last resort. tricky for vulnerable people
- right to withdraw without penalty and full compensation
debriefing
end or research - provide info, undo any harm, inform about deception
Scientific fraud =
1.
2.
3.
4.
data falsification / manipulation
1. coding done wrong
2. make up data
3. alter data
4. create co-author
questionable research practices
1.
2.
3.
4.
1.optional stopping
2. dropping some dependent variable
3. dropping conditions
4. using covariates self-servingly
QRP solution
report and register data collecting rules in advance
____ is the difference between scientific ___ and ____, but not becoming more ___
intent, fraud, QRP, alike
____ decisions are problematic
self-serving
HARKing
hypothesizing after results are known
Generalizability =
1.
2.
3.
how to interpret results
1. what population does this sample represent
2. in what contexts would we see similar results
3. are these results repeatable