Association claims are assessed via
Bivariate correlations
Bivariate correlation
An association between scores on two measured variables
X=score on variable A
Y= score on variable B
Both at least interval
Persons R
Scatter plot
Ordinal and one at least ordinal
Spearman’s rho
Scatter plot or bar
Categorical one at least ordinal
T-test (or equivalent non parametric test) 2 GROUPS
ANOVA (or equivalent non parametric test) 3 OR MORE GROUPS
Bar graph
Persons R
-1 to 1
Says strength and direction
.1 small or weak
.3 medium or moderate
.5 large or strong
If it comes first in time it is on what axis
X axis
Under what conditions should you construct a scatter plot
Both are at least interval
Under what conditions should you construct a bar
One categorical, at least ordinal
One ordinal, at least interval
What information can a scatterplot provide about the relation between two variables
Shape direction and strength
Which forms of validity are relevant when interrogating association claims
Construct
Statistical
External
Questions to assess construct validity
How well was each variable measured?
Do the measuring have good reliability?
Are you measuring what it’s intended to measure?
Questions to assess statisical validity
What is the effect size?
Is the association curvilinear?
Could there be outliers be affecting the association?
Is there restrictions of range?
Questions to assess external validity
What sampling technique was used?
Can you generalize to outside populations?
Does the context of the study represent real world connections?
Does it hold true in different time place and people?
Effect size
The magnitude of a result
How strong your association is?
Descriptive, the numbers describe the data
Persons R
Cohens D
0 to infinity
One is categorical/ ordinal other is at least interval
0-.2 small
.21-.79 medium
.80+ large
What do you use if you want to make an inference
Null hypothesis significance testing
Is your correlation is significant?
Are group differences significant?
Three steps
NHST
Effect size
Practically relevant
Within group
Dependent group design
Everyone goes through every condition
Minimize variability more internal validity
Pay attention to order of trials
Between groups design
Independent
One group per condition
Pseudo random assignment
Want the groups to look similar based off characteristics still random semi like stratified random sampling
Control of variables that could effect your measurement
Matched pairs design
Match people on some variable that isn’t the IV but could effect the DV
Pairs get split up
