sample vs. population
sample: subset of people taken from the population used to make inferences
population: the entire set of people that you want to draw conclusions from
correlational studies support:
association claims
in a correlational study, all variables are:
measured
goals of correlational research
pattern prediction, supporting reliability/validity of a measure, preliminary evidence to determine if experiment is worth conducting
how to evaluate statistical validity of a correlational study
effect size (r): strength of relationship between 2 variables
confidence interval: range of values
statistical significance: p-value
range of effect size (r)
small effect: r=0.10
moderate effect: r(0.1, 0.5)
large effect: r>0.5
what criteria of causality is NOT met by a correlational study?
internal validity: cannot control for 3rd variables
[temporal precedence – only if the correlation isn’t longitudinal]
issues when assessing correlation vs causation
restriction of range: sample doesn’t capture full range of variables
construct validity: how well the variables were measured
external validity: are the results generalizable?
Assumptions of Pearson’s r-coefficient
levels of measurement: both variables continuous
related pairs: no missing data
absence of outliers
linearity: line of best fit should be linear
purpose of Pearson’s r-coefficient
correlational coefficient. Tells you how strongly 2 variables are correlated.
how to report correlations
r(df) = .xx, p = .xxx, 95% CI [.xx, .xx]
There was a significant correlation between number of publications and salary (r(60) = .65, p < .001, 95% CI [.48, .77]) such that having more publications was associated with having a higher salary.
