Methods of Meta-Analysis: Correcting Error and Bias in Research Findings by Frank L. (Leo) Schmidt & John E. Hunter

Methods of Meta-Analysis: Correcting Error and Bias in Research Findings by Frank L. (Leo) Schmidt & John E. Hunter

Author:Frank L. (Leo) Schmidt & John E. Hunter [Schmidt, Frank L. (Leo)]
Language: eng
Format: epub
ISBN: 9781483324517
Publisher: SAGE Publications
Published: 2014-02-27T00:00:00+00:00

Exercise 7.1: Meta-Analysis of d Values

A hypothetical researcher conducted a review of experimental studies of gender differences in conformity. The 10 studies summarized here are called “other conformity studies.” Because all these studies use an experimental paradigm involving a nonexistent norm group, they are called “fictitious norm group” studies.

These studies measure conformity by examining the effect of knowledge of other people’s responses on an individual’s response. Typically, an experimental subject is presented with an opportunity to respond to a question of opinion. Before responding, the individual is shown some “data” on the responses of other individuals. The “data” are manipulated by the experimenters, and the “other individuals” are the fictitious norm group. For example, the subject might be asked for an opinion on a work of art and told that 75% of art majors liked the work “a great deal.”

Positive d values indicate females were more conforming; negative values indicate males were more conforming.

Except for Step 2, this exercise can be worked using the program D-VALUE, which corrects d values individually for artifacts. Alternatively, you can convert all d values to correlations and apply the program VG6, which uses the methods presented in Chapter 3 to correct correlations individually and then performs meta-analysis on the corrected correlations. You would then convert these results back into the d value metric using equations given in this chapter. Both these programs are included in the Windows-based package of programs available for applying the methods presented in this book. Details about this software package can be found in the Appendix. However, you will learn more if you carry out the exercise using a calculator or a spreadsheet.


1. Perform a bare-bones meta-analysis of these 10 studies. Correct the d values for their small positive bias, as discussed in this chapter. In estimating sampling error variance, use the most accurate formula given in this chapter. Create a table showing the mean d, the observed variance of d, the variance predicted from sampling error, the variance corrected for sampling error, and the standard deviation of the d values corrected for sampling error. In addition, present the percentage variance accounted for by sampling error variance, the square root of the proportion of variance accounted for, and the 80% credibility interval.

What percentage of variance is accounted for by sampling error? What is the correlation between sampling errors and the observed d values? How would these results be interpreted by someone with no knowledge of measurement error? That is, give the “face value” interpretation of these results.

From these results, in what percentage of groups would you expect males to be more conforming? Females? Note that if were 0, males would be more conforming in 50% of the groups and the same for females. Use the properties of the normal curve to compute the percentages implied by your value. First, compute . Then look this z value up in a normal curve table and record the percentages above and below this z value. These are the needed percentages.

2. Determine the importance of correcting the observed d values for their positive bias.


Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.