Week 11 Effect Size and Power Flashcards
What is omega squared?
a measure of treatment magnitiude
What is omega squared based on when applied to a single factor design?
Variances:
- Difference among population treatment means
- variability within treatment populations.
What does effect size measure?
The approximate percentage of variance that is due to experimental treatment
What are the properties of Omega Squared?
- provide relative measure of strength of IV
- logically can range from .00-1.00
- empirically more restricted
What are the terms to describe the size of an effect?
Small - .01
Medium - .06
Large - .15
In an F test, if F is significant, what is the omega squared?
Omega squared is significantly > 0
Is omega squared effected by small sample sizes?
No even when non-sig F
What is R-squared?
Squared multiple-correlation coefficient. Always > than omega squared
What does power reflect?
- Degree to which one can detect treatment differences
- Changes others will be able to duplicate findings
What can power do?
- Control magnitude of type 1 error (1-a, true Ho, rejection)
- Control type 2 error (1-B, false Ho accept)
What is type 1 error due to?
Significance level (.05)
What is type 2 error due to?
- Size of treatment effects
- Sample size
- Degree of error variance
- significance level
What are some ways to increase power?
- Strong manipulations
- Increase sample size
- Increase probability of type 1 error rate
- Reduce error variance
What is the sensitivity of an experiment related to?
Size of error component
What are the three major sources of error variance?
- Random variation in actual treatments
- Unanalyzed control factors
- Individual differences
