Class Notes (1,100,000)
US (470,000)
UMN (4,000)
KIN (20)
Alex K (10)
Lecture 13

KIN 3982 Lecture Notes - Lecture 13: Meta-Analysis, Sampling Bias

Course Code
KIN 3982
Alex K

This preview shows half of the first page. to view the full 2 pages of the document.
3/27 Notes
Wednesday, March 27, 2019
2:39 PM
What effect size values are good?
When "d" is calculated:
o Small = .20
o Medium = .50
o Large = .80
o Very large = 1.10
o Extremely large = 1.4+
Maximum is 3.0 but you rarely see values over 1.0 in the social science
Interpreting Effect Size
It is possible for a small effect size to represent an important result
o Helping ill individuals by using a new treatment
It is possible for a large effect size to represent unimportant results
o If the results lack practical significance regarding cost, public
acceptability, and ethical concerns, then could be unimportant
Three steps for interpreting results
o P-value less than .05
o Look at effect size (d)
o Consider in context of theory and practical significance
Other considerations when interpreting results
Inadequate sampling
o E.g., very biased sample
Invalid instrumentation
o Low reliability and validity
Poor research design
For the above examples, the effect size could be seen as meaningless
Traditional subjective narrative review vs. meta-analysis
o Traditional narrative review - the authors summarize the findings
"three of the four studies found an effect"
o Meta-analysis - set of statistical methods for combining the results of
the studies
The authors average the results by summing the mean differences
and dividing by the total number of means (or studies)
You're Reading a Preview

Unlock to view full version