Statistical Validity for Associations: Effect Sizes, Precision, and Statistical Artifacts

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/7

encourage image

There's no tags or description

Looks like no tags are added yet.

8 Terms

1
New cards

Six Core Questions of Statistical Validity for Association

  1. effect size

  2. confidence intervals

  3. replicability

  4. outliers

  5. restriction of range

  6. curvilinear associations


2
New cards
  1. How strong is the relationship (effect size)


represents the magnitude or strength of the relationship between variables

3
New cards
  1. 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

4
New cards

How do sample size affect precision

small samples : estimates bounce around → very wide

large samples : estimates settle down and intervals get narrow

5
New cards
  1. 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


6
New cards
  1. Is there restriction of range


sample does not encompass the full spectrum of variation present in the broader population for one or both variables

7
New cards
  1. Is the relationship curvilinear


a relationship between two variables cannot be captured by a straight regression line

8
New cards
  1. Has the finding been replicated


repeating an empirical study using identical/ equivalent methods to evaluate if findings are consistent