Exam 3 Flashcards
(64 cards)
1
Q
Multivariate designs
A
- involve more than 2 measured variables
2
Q
3 criteria for causation (apply these to correlation research)
A
- covariance
- temporal precedence
- internal validity
3
Q
longitudinal design
A
- can provide evidence for temporal precedence by measuring the SAME variable in the SAME people at several points in time
- used in developmental psychology
- same variable, same group, over time
4
Q
Results of a longitudinal design (3 correlation types)
A
- because multiple variables involved -> gives individual correlations (3)
- 1 cross-sectional correlation = test to see whether 2 variables, measured at the same points in time are correlated
(over evaluation time 1)-> (narcissism time 1) - cannot alone establish temporal precedence
2. Autocorrelations - the correlation of one variable with itself, measured at 2 different times (overvaluation time 1)-> (overvaluation time 2).
3. cross lag correlation - show whether the earlier measure of one variable is associated with the later measure of the other variable (3 results)
diagonal correlations
5
Q
longitudinal studies can provide some evidence for a causal relationship
A
- covariance - when 2 variables correlated and CI does no include ) - covariance
- temporal precedence -> each variable measured at clearly different points
- internal validity -> when only measuring 2 key variables , longitudinal studies cannot rule out 3rd variables
6
Q
multiple regression -> deals with internal validity
A
- a statistical technique that computes the relationship bw a predictor variable and a criterion variable/controlling for other predictor variables
7
Q
control for
A
holding a potential 3rd variable at a constant
- accounting for subgroups
8
Q
2 variables
A
- criterion variable = dependent variable, variable researchers are most interested in understanding or predicting
- predictor variables - indep. variables, used to explain variance in the criterion variable
9
Q
Beta
A
- similar to r
positive beta - positive relationship between predictor and criterion
negative - negative relationship b/w predictor and criterion variable
beta that is zero - no relationship
higher beta = stronger
lower beta = weaker
10
Q
common phrases with regression in media
A
” controlled for”
- “adjusting for”
- “considering”
11
Q
regression does not equal causation
A
- even though multivariate designs
- analyzed w regression stats can rule out 3rd variable
- can’t establish temporal precedence
- well run experiments more convincing then causation
12
Q
parsimony
A
- degree to which a scientific theory provides the simplist explanation of some phenomena
- ex - simplest context of investigating data
13
Q
mediator/mediating variable
A
- a variable that helps explain the relationship between two variables, can be experimental or correlation study
- mediation hypothesis -> causal claims
- must have temporal precedence
14
Q
mediators vs 3rd variables
A
mediators
- both have to be tested
- tells theoretical meaning
- direct interest to the researcher
- why are variables linked?
3rd variables
- variable is external to the original bivariate correlation
- nuance
BOTH CAN BE TESTED W/multiple regressions
15
Q
4 big validity questions for multivariate designs?
A
= same as bivariate correlation
1. internal
2. construct
3. statistical
4. external
16
Q
Interaction effect
A
- a result fro a factorial design in which the difference in the levels of the IV variable changes, depending on the level of the other IV, a difference of differences also called interaction
17
Q
Interaction
A
the difference in differences, the effect of one IV depende on the level of the other IV
18
Q
factorial design
A
- a study in which there are 2 or more IV or factor, way researcher tests for interactions
most common -> researchers cross the 2 IVs, study each possible combo cell, one of the possible combos of 2 IVS
19
Q
simplest factorial design (2 IVS/FACTORS)
A
- cell phone use, age (each has 2 factors)
- 2x2 factorial design
- the levels of the IV are crossed w/2 levels of the other IV
2 x 2 = 4 cell design
20
Q
participant variable
A
- a variable whose levels are selected (measured)
- not manipulated
ex - gender, ethnicitu,
21
Q
external validity
A
- testing limits
- when researchers test an IV in more than one group at once, testing whether the effect generalizes
22
Q
moderators in factorial design
A
- moderator is an IV that changes the relationship between another IV and a DV
For whom does variable a cause variable b for what situations does variable a cause variable b
23
Q
interpreting factorial results
A
1 main effect
- overall effect of one IV on the DV averaging over the levels of the other iV
- main effect -> simple difference
- factorial design w/2 IVs = main effects
24
Q
marginal means
A
- arithmetic means for each level of the IV averaging over other levels of the IV
- sample size is equal: marginal means are a simple average
- sample size is unequal - marginal mean computed using the weighted average counting the larger sample more
25
estimating how large each main effect is
- calculate difference between marginal means
- compute 95% CI
- contains 0 - not significant
- doesn't contain 0 = significant
26
interaction effect
the difference in differences
detecting on a graph
- line graph - lines are not parallel may be interaction
- lines are parrallelish- no interaction
- interactions are more important than main effects
27
factorial design variations
- independent groups factorial design
- both IVs are studied as indep. groups
- if a design is 2x2 (4 groups)
within groups factorial design
- both IV variables manipulated as within groups
- if design is 2x2, one group participates in all 4 combos, cells of design
counterbalanced
- fewer participants
mixed factorial design
- one IV is manipulated as Indep. groups, other is manipulated as within groups
- levels of IV can increase
- can have 3 levels 2x3 = 6 design
28
increasing the number of the iV's
- 2x2x2 factorial design
- three way design
- 3 main effects
- 3 possible two way interactions
29
identifying factorial design in popular media
- "it depends"
- "only when"
- look for participant variable (age, gender)
30
External validity
generalize to other samples + settings, focuses on samples
31
population
Sample -->
- entire set of people or products in which you are interested
sample = smaller set, taken from that population
32
representative sample
- unbiased allow us to make inferences
- all members of the pop. have an equal chance of being included in the sample
33
biased sample
- unrepresentative sample
- some members of the population of interest have a much higher probability than others of being included in the sample
34
ways a sample can be biased
1. sampling only those who are easy to contact - convenience sampling
2. sampling only those who volunteer
- self selection
- ex - self selecting, rate my professor leaving review on amazon
- ppl who rate things may have stronger opinions, more likely to share ideas
35
probability sampling/ random sampling
- best way to get a representative sample
- every member of the pop of interest has an equal and known chance of being selected for the sample, whether they are convenient or motivated by volunteer
- excellent external validity, can generalize to the pop of interest
36
random sampling - simple random sampling
- assign a number to each individuals, select random numbers
- computerized randomizer
37
random sampling - systematic sampling
- researchers use a random number table, computer
- selecting 2 random numbers (4 and 7)
- start with/fourth person count off by 7
- suprisingly difficult and tine consuming
38
random sampling - cluster sampling
- ppl already divided into arbitrary groups clusters of participants within a pop. of interest randomly selected individuals in each of the clusters are used
- list of HS in PA, select 100, include every student from 100
39
random sampling - multistage sampling
- two random samples are selected at a random sample of clusters and then a random sample of ppl within those clusters
ex = HS PA, random 100, random sample of students from each of the randomly selected schools
40
random sampling - stratified random sampling
- researcher purposefully selects particular demographic categories or strata, and then randomly selects individuals within each group, proportionate to there assumed membership in the population
- stratas (meaningful categories) -> ethnic, religious groups
41
random sampling - oversampling
- researcher intentionally overepersents one or more groups
- adjust final result so members of oversampled group are weighted to there actual proportion in the pop.
42
random assignment
- used only in experimental designs
- when researcher want to place participants in 2 diff groups, they randomly assign
43
non probability sample technique - connivence sampling
-sample from a group which is accesible, psych professor using psych students
44
non probability sample technique - purposive
- researchers want to study only certain kinds of people, recruit only those participants
ex - smoke, research smokers
45
non probability sample technique - snowball sampling
- participants are asked to recommend a few acquaintances for the study
- ex = people who have crohn's disease ask them to recruit people from a support group
46
non probability sample - quota sampling
- similar to stratified sampling, researcher identifies subsets of the pop. of interest and then sets a target number (quota) for each catagory
47
external validity
- bigger samples are not always better
- larger samples not more externally valid
- larger the sample -> smaller the margin of error
- sample size not an external validity isse, statisitical validity issue
48
quasi experiment
- not able to assign all. participants to one level or another
49
quasi independent variable
because the researcher didn't havre full control over the IV
50
4 types of quasi experiments
1. nonequivelant control group post test only design
2. nonequivalent control group pretest/postest design
3. interrupted time series design
4.. nonequivalent control group interrupted time series design
51
quasi experiment type 1 - nonequivelant control group post test only design
- participants were not randomly assigned to groups and were tested only once after exposure to one level of the IV or the other
52
quasi experiment type 2 - nonequivelant control group pretest/postest design
- participants were not randomly assigned to groups and were tested both before and after surgery
53
quasi experiment type 3 - interrupted time series design
- measures a variable repeatedly before, during and after the "interruption" caused by some event
- example suicide rate, the show 13 reasons why
54
quasi experiment type 4 - nonequivalent control group interrupted time series design
- combines 2 of the previous designs
- 2 groups were not randomly assigned to having the pill group (nonequivalent groups) and researchers did not have control over when the year law passed (interrupted time series)
55
what is the main concern with quasi experiments?
- internal validity
56
how to control for selection effects in a quasi design?
- waitlist design
- all participants plan to receive treatment but are randomly assigned to do so at different times
57
design confounds in quasi design
- design confounds - something else systematically varies
- maturation threat
- history threat
- regression to the mean
- attrition threat
- testing and instrumentation threats
- observer bias, demand characteristics and placebo effect
58
What does a researcher gain from a quasi experiment?
- real world opportunities ex - opioid crisis, cosmetic surgery
- external validity
- real world settings of many quasi experiments can increase likelihood of external validity
- ethics
- many questions to researchers would be unethical to study
59
quasi experiments and construct validity
- how successfully the study manipulated or measured variables
60
quasi experiment and statistical validity
- how large groups differences (effect size)
- CI
61
quasi indep variable vs partcipant variable
quasi indep. variable -focus on potential intervention
participant variable - categorical, document similarities, differences, indiv. differences
62
stable baseline design
- researcher observes behavior for an extended baseline period before beginning a treatment or other interventions
- behavior during baseline is stable, more certain treatment is effective
- stable baseline -> internal validity
63
multiple baseline design
- researchers stagger there introduction of an intervention across a variety of indiv. times, or situations to rule out alternative explanations
ex - 3 children starting 3 different times
64
reversal design
- researcher observes a problem behavior both before and during treatments, and then discontinues treatment for a while to see if behavior returns
internal validity -> observing how behavior changes as treatment is removed
