Maastricht University
Question
Asked 28 May 2016
Meta-Analysis: Use Number of Observations or Number of Individuals/Firms?
Hey,
I am computing a Meta-Analysis to find out whether Family Firms or Non-Family Firms are more successful.
Unfortunately, most primary studies only report panel data. They report the Number of Firms AND the Number of firm-year observations.
Which one of these should I use as my sample size? My goal is to compute the mean correlation effect size.
One study (van Essen, M., Carney, M., Gedajlovic, E. R. and Heugens, P. P. M. A. R. (2015), How does Family Control Influence Firm Strategy and Performance? A Meta-Analysis of US Publicly Listed Firms. Corporate Governance: An International Review, 23: 3–24. doi: 10.1111/corg.12080) obviously used the number of firm-year observations. Another (E.H. O’Boyle Jr., J.M. Pollack, M.W. Rutherford: Exploring the relation between family involvement and firms’ financial performance: A meta-analysis of main and moderator effects
Journal of Business Venturing, 27 (1) (2012), pp. 1–18) used the number of firms.
Thanks in advance.
Todor
All Answers (3)
Dear Todor,
Considering that the original articles already controlled for the dependency between data points from the same firm, I think you can safely assume the number of firm-year observations as the sample size. Taking the number of firms will underestimate the number of actual observations the effect size is based on, and thus will receive too little weight in the meta-analysis.
1 Recommendation
Vanderbilt University
Hi Todor,
You want to remain consistent across all your studies to have an interpretable overall mean effect size in the end. If you have multiple effect sizes of interest from the same study, you will want to control for the within-study dependence via robust variance estimation. In terms of weighting, the meta-analysis you conduct (e.g., random-effects meta-analysis) should account for the individual-study sample size by weighting the effect size using its variance/SE. Depending on how many studies you have, you can theoretically propose to control for other study-level factors in meta-regression analyses. Goodluck!
1 Recommendation
Similar questions and discussions
Proportional similarity index PSi: sensitivity to number of diet items?
I have a question about how to interpret (differences in) the proportional similarity index Psi proposed by Bolnick et al. (2003) https://search.r-project.org/CRAN/refmans/RInSp/html/PSicalc.html
The question centers on the sensitivity of the index to the number of ‘diet items” that are used by individuals.
I am testing for differences in diet for individuals in a single population, either using ‘ingredients’ as input, or using derived ‘nutrients’ as input. For ‘ingredients’, there are up to 20 different items an individual can consume [e.g. ants, different berries and grains, egg, meat, etc] while ‘nutrients’ are limited to about 5 categories [protein, fat, carbohydrates, fiber, ash].
Are the resulting PSi values comparable? Say that an individual scores Psi = 0.1 based on ingredients but the corresponding nutrient-based Psi = 0.8, can I conclude that an individual acts as a specialist when it comes to ingredients but as a generalist for nutrients?
Or are these scores completely incomparable because of the different number of ‘items’ they are calculated on?
Thanks for any advice,
Diederik
Related Publications
High-dimensional low sample size (HDLSS) data sets emerge frequently in many biomedical applications. A common task for analyzing HDLSS data is to assign data to the correct class using a classifier. Classifiers which use two labels and a linear combination of features are known as binary linear classifiers. The direction-projection-permutation (Di...
Meta-analysis is a set of statistical procedures used for providing transparent, objective, and replicable summaries of research findings. This tutorial demonstrates the most common procedures on conducting a meta-analysis using the R statistical software program. It begins with an introduction to meta-analysis along with detailing the preliminary...