exam 1 Flashcards
3 features of an experiment
- vary at least one or more independent variable (s)
- assign participants to experimental conditions in a way that ensures their equivalence
- experimental control
how do we know if an iv is strong enough to test/ has been manipulated enough?
pilot test: little version of actual experiment to get a sense of how things are going ; great for finding mistakes
manipulation check: question that you ask participants to see if independent variable had an effect
how do groups get roughly equivalent groups for each condition?
self selection: participants choose condition themselves (bad bc diff types of ppl in conditions)
arbitrary assignment: assignment based on a rule (could still put diff types of ppl in conditions)
matching on multiple variables: each participant has a ‘twin’ in other condition (hard to do)
random assignment: putting them in group based on random number (best way to ensure equivalence)
what are the two types of experimental design?
between subjects design: each participant is tested under only one condition/level
within subjects design: each participant is tested under every condition / level of IV (repeated measures)
characteristics of a within subject design
- looks like pretest / post test design
- pros: more powerful (ability to detect differenes / effects in group) , fewer ppl needed
- cons: order effects (practice, fatigue), sensitization (figuring out what study is about), carryover (effect in 1 condition lingers onto next)
how to minimize order effects
counter balancing: diff participants complete levels of IV in diff orders
- latin sqaure design: method of counterbalancing where all possible orders are used
within vs between subjects design when…
within : order effects are not a problem, powerful design is wanted, there are scarce participants, you want to generalize, participants are exposed to multiple levels of treatment
between: order effects are a problem, there is large # of participants, in real life people receive one not both levels / conditions
what is experimental contol ?
experimental contol: hold extraneous varibles constant
systematic variance vs error variance vs confound
systematic variance: variance across experimental groups (differences between groups)
error variance: is everything we’re not looking for (variance between groups)
confound: variable other than IV differes between groups (systematic)
- treatmment + confound = systematic
internal validity
degree to which a researcher can draw accurate conclusions about the effects of an IV
- relies on elimination of confounds
-achieved through experimental control
what is biased assignment?
threat to experimental control
people are put into groups in biased / non random way (between subjects)
what is differential attrition?
threat to experimental control
‘mortality’ / people dropping out of a study (within & between subjects)
only a confound if one group has drop outs and other does not
what is pretest sensitization?
threat to experimental control
when taking pretest changes you
- participant reats differently to IV than they would have if not presented with pretest (within)
what is ‘history’?
threat to experimental control
anything happening outside of study that might account for the outcome (within & between)
-local history effect: something outside is influencing one group but not the other
- how to reduce: shorter time inverals btwn pre and post test & replicate studies
what is maturation?
threat to experimental control
peoples natural development (within)
- internal changes in the participant during the experiment
what is regression to the mean?
threat to experimental control
if one sample of is extreme, the next sampling is likely to be closer to its mean (within)
- dont test extreme groups if possible
-have control group thats equally extreme
what is instrumentation?
threat to experimental control
changes in the measurement tool across conditions or testings (within & btwn)
- make sure instruments are calibrate
what is diffusion?
threat to experimental control
participants in diff conditions communicate w/ each other (btwn)
- ruins ability to detect IV
what is lack of standardization?
threat to experimental control
there is no standardized procedure: scripts, trained experimenters, same: time, temp, location, lighting, noise level
what is design confounds?
threat to experimental control
order effects could be confounds (within)
what is experimenter expectancy effects?
threat to experimental control driven by people involved
the researchers expectations about the outcome of a study influences participants reactions
- keep experiment blind
what is demand characteristics?
threat to experimental control driven by people involved
participants catch onto point of study and change their behavior to how they think they should behave
what is social desirability / evaluation appehension?
threat to experimental control driven by people involved
just want to look good, not giving honest responses to portray in a positive light
- make anonymous
-“be honest”
what is the placebo effect?
threat to experimental control driven by people involved
when someone experiences a change because they expect a change
what is the Hawthorne effect?
threat to experimental control driven by people involved
actualy knowing that you are in an experiment leads to diff outcomes than would normally occur
-have both groups think they’re in the study or not
what is error noise?
static in the measurement
- messes up both groups
- “waters down the experiment” =n effect is not as strong
- related to reliability
what is external validity?
degree to which results obtained in one study can be replicated or generalized
what is “across situations” in external validity ?
can results of study be generalized in a different study ?
- ecological validity: generalized to real world
- mundane realism: experiment is similar to real life situations in a lab setting
- psychological realism: psychological processes are similar to those in real life situations (does it FEEL real?)
what is “across people” in external validity ?
do people in sample generalize to a more broad population?
- random selection of participants is best
-most common psych participants fall into: Western Educated Industrialized Rich Democratic
what is “across operationalization” in external validity?
same outcome if we had measured our conceptual variable differently
t statistic
estimated standard error of population based on sample
- the denominator is always diff bc sample variance changes from sample to sample unlike pop variance that is same for all
why is variance used instead of standard deviation
standard deviation is descriptive and biased vs variance which is inferential and unbiased
degrees of freedom
n-1
- the bigger the df, the better the sample, bigger sample better reflects the population
t disributions
- smaller distributions (smaller sample sizes) have more variability / flatten out
- as sample size gets bigger, distribution is closer to normal distribution
- shape of t distribution changes with df
what does numerator in one sample t test represent?
actual difference between pop and sample mean
what does denominator in one sample t test represent?
average (expected) of how sample differs from pop mean
assumptions of t test
-independent observations
- population sampled must be normal
advantages of a t test
population mean and pop standard error are not needed
independent measures / btwn subjects is..
design involves 2 seperate and independent samples and makes a comparison btwn 2 groups
repeated measures / within subjects is…
two sets of data obtained from same sample
what does the numerator in a t test for independent samples mean?
sample mean diff - pop mean diff
what does the denominator in a t test for independent samples mean?
standard error of sample mean difference
standard error
amount of error expected when you use sample mean diff to estimate pop mean difference (how well sample represents pop)
assumptions of independent t tests
- observations are independent
- 2 population samples are selected from are normally distributed
- homogeneity of variance: two populations the samples are selected from must have same variance (fmax should not be sig for there to be homogeneity of variance)
the two types of related measures t test are:
repeated measures: single sample of individuals is measured more than once on the same dependent variable
matched subject study: each individual in one sample is matched on a specific variable with a subject in another sample
there is less variance/ less standard error/ bigger t bc we’re comparing ppl to themselves
big diff betwn related t test and others
comparison distribution is no longer just a sample mean; comparison distribution is now mean difference score
what is difference scores?
value obtained from subtracting the before treatment score from after treatment score (x2-x1)
assumptions of related measures t test
-observations must be independent
- population distribution of d scores must be normal
