1/7
Looks like no tags are added yet.
Six Core Questions of Statistical Validity for Association
effect size
confidence intervals
replicability
outliers
restriction of range
curvilinear associations
How strong is the relationship (effect size)
represents the magnitude or strength of the relationship between variables
How precise is the estimate (confidence intervals)
sample correlation (r ) : single guess at the true relationship in the whole population
confidence interval (CI) : range of values where the true relationship falls
95% CI does not contain zero (p<.05) = statistically significant
95% CI contains a zero (p>=.05) = non-significant
precision : narrow interval = precise estimate
How do sample size affect precision
small samples : estimates bounce around → very wide
large samples : estimates settle down and intervals get narrow
Could outliers be distorting the association
outliers are extreme score that deviates from the general distribution of data points
online outliers : along the trajectory of the regression line, inflate the magnitude of r
offline outliers : far off the regression line, attenuate the magnitude of r
Is there restriction of range
sample does not encompass the full spectrum of variation present in the broader population for one or both variables
Is the relationship curvilinear
a relationship between two variables cannot be captured by a straight regression line
Has the finding been replicated
repeating an empirical study using identical/ equivalent methods to evaluate if findings are consistent