Exam 3 multivariate Flashcards
multivariate designs
involve more than two measured variables
* correlational research can easily establish covariance
* longitudinal designs help address temporal precedence
* multiple regression analyses help address internal validity
longitudinal designs
can provide evidence for temporal
precedence by measuring the same variables in the same people at several different times
* used by developmental psychologists
* gets us closer to a causal claim
what are the 3 kinds of longitudinal designs
cross-sectional correlations, autocorrelations, and cross-lag correlations
what is an example of a longitudinal design
Measuring two variables (parental overvaluation, child’s narcissism)
at four different time points.
Overevaluation time 1 and Narcissism time 1
Overevaluation time 2 and
Narcissism time 2
Overevaluation time 3 and
Narcissism time 3
Overevaluation time 4 and
Narcissism time 4
cross-sectional correlations
cross- sectional correlations test the relationship between two variables measured at the same point in time
Overevaluation time 1 and Narcissism time 1 r=0.007
Overevaluation time 2 and
Narcissism time 2 r=0.70
Overevaluation time 3 and
Narcissism time 3 r=0.138
Overevaluation time 4 and
Narcissism time 4 r=0.99
autocorrelations
autocorrelations test the relationship a variable and itself across time
Overevaluation time 1 ->0.695 Overevaluation time 2 ->0.603
Overevaluation time 3 ->0.608
Overevaluation time 4
cross lag correlations
cross-lag correlations test the relationship between a variable at time 1with another variable at time 2 -> à temporal precedence
in a multivariate design why not do an experiment?
In many cases participants cannot be randomly assigned to a variable
(e.g., parenting style)
People cannot be assigned to preferences
(e.g., favorite kind of music)
It may be unethical
(e.g., making children narcissistic or depressed)
multiple regression analyses
when you measure more than two variables, you can check whether a third variable affects the main association of interest
– is the relationship between social media use and depression simply due to the participants’ age?
– you can “control” for the age variable to find out
- a statistical method of checking for “problematic third variables”
and “spurious correlations”
– instead of simply looking at a scatter plot, you can run a test
to get a numerical, statistical answer
how can you use statistics to control for third variables
Regression results indicate whether a third variable explains the relationship
* criterion variables (dependent)
* predictor variables (independent)
* interpreting beta
popular regression phrases in media
controlled for
adjusted for
accounting for
considering
does regression establish causation?
no multiple regression is not a foolproof way to establish causation. There is always something more a researcher could have controlled for: it’s not possible to control for everything. Regression does not establish
temporal precedence
pattern and parsimony
simplest explanation requiring the fewest assumptions
mediation
asks why
ex: Depression is related to social media use. Why?
– Maybe depression decreases time in in-person activities, increasing amount of phone time.
Ex:Amount of deep talk is related to well-being. Why?
– Maybe deep talk increases the quality of social ties.
moderators
ask for whom or when
ex: Depression is related to social media use. For whom?
– There is a stronger link for girls than boys (gender is a moderator)
ex:Amount of deep talk is related to well-being. In what situations ?
– Researchers found that weekday/weekend conversations had the same relationship with well-being (day of week was not a moderator
similarities and differences between mediators and problematic third variables
Similarities: both involve multivariate research designs, and both can be detected using multiple regression.
Differences: third variables are external to the bivariate correlation (problematic), whereas mediators are internal to the causal variable (not problematic).
