Meta-analysis

Process of averaging over studies to find results, patterns and interesting relationships among variables.

Steps of meta-analysis

1.) Finding the studies

2.) Quality screening?

3.) Find common measure of effect size for all the sudies

4.) Identify moderator variables

5.) Average studies to get overall measure

6.) Examine effects of moderator variables

2.) Quality screening?

3.) Find common measure of effect size for all the sudies

4.) Identify moderator variables

5.) Average studies to get overall measure

6.) Examine effects of moderator variables

Problems with finding all of the studies on a subject

- Very difficult. Have to use many different search engines and even still that only gets the published ones (many research articles aren't published)

- Published articles may be biased in that publishers tend to report studies that have found an effect (and not those that haven't or are sitting on the fence).

- Published articles may be biased in that publishers tend to report studies that have found an effect (and not those that haven't or are sitting on the fence).

Meta-analysis tends to get a result that is an..

Overestimate of the real world effects. This is because many studies that find no result are simply not published.

Better option than quality screening

CODING system and use the quality code as a moderating variable (QS Increases the bias)

Methods for making the different studies dependent and independent variables (which are often very different) comparable

- Correlations between the two variables of interest for each study. (eg. job performance and commitment)

- Take the difference in the means and divide by the standard deviations (d). Done when the independent variable is a two category one (e.g. reward someone or you don't)

- Take the difference in the means and divide by the standard deviations (d). Done when the independent variable is a two category one (e.g. reward someone or you don't)

Things we average to get the overall effect size

- How many subjects they use

- Correct for reliability (unreliable/inaccurate measures bias the end result downwards)

- Correct for reliability (unreliable/inaccurate measures bias the end result downwards)

Process of examining effects of moderator variables

Similar to doing a t-test or one way ANOVA.

- Effect size is dependent variable and coded moderator is the independent variable

- Effect size is dependent variable and coded moderator is the independent variable

d=

(means rewarding - means doing nothing)/SD