Lecture Conducting and reading research in health and human performance (4/e): Chapter 11 - Ted A. Baumgartner, Larry D. Hensley

Chapter 11
Meta­Analysis


Meta­analysis





Quantitative means of reanalyzing the
results from a large number of research
studies in an attempt to synthesize findings
More than merely a review of related
literature
Relatively new approach in HHP research


Effect Size





Basic statistic used in meta-analysis
Converts results from different studies to a
common metric so that comparisons can
be made
Used to estimate meaningfulness of an
outcome (i.e., practical significance)
– Not influenced by sample size

Example Size Formula
ES = (Me – Mc)/Sc

Formula for estimating ES for difference between experimental
and control group. Where Me is the mean of the experimental
group, Mc is the mean of the control group, and Sc is the
standard deviation of the control group.

Note: There are numerous formulas that can be used to calculate ES.


Interpretation of ES
< .20 small
.50 medium
> .80 large

Source: Cohen (1988)


Meta ­Analysis


In meta-analysis, each research study
contributes a data point to the subsequent
analysis, much like an individual
participant in a descriptive or experimental
research study

Steps in Meta­Analysis
1.

Compile references
•
•

1.

Determine inclusive criteria
•

1.

There must be a substantial number of
research studies available on a topic
Requires means, standard deviations,
correlations, etc. be published
E.g., published in last 10 years or N > 30

Review each study
•
•

Record information needed to calculate ES
from each study
Identify and code moderator variables, if any

Steps in Meta­Analysis cont.
4.
5.

Decide which studies to use
Do the meta-analysis
•
•
•
•

4.

Calculate the effect size for each study
Generate summary statistics for effect sizes
Examine according to moderator variables
Interpret results

Report the results


Criticism of Meta­Analysis





Not the ultimate answer
Does not differentiate in quality of studies
Combines unlike studies with too much

variability (i.e., mixing apples and oranges)
Inappropriate coding of variables