Meta-Regression-Based Co-Word Analysis
Also known as: MR-CWA, meta-regression co-word mapping, regression-weighted co-word analysis, co-word meta-regression
Meta-regression-based co-word analysis is a hybrid scientometric technique that enriches traditional co-word mapping by weighting keyword co-occurrence networks with meta-regression-derived effect estimates. Instead of treating all documents as equally informative, the method uses statistical regression to incorporate study-level moderators — such as publication year, sample size, or methodological quality — into the co-occurrence structure, revealing how thematic clusters in a research field vary across moderator conditions.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use meta-regression-based co-word analysis when you are conducting a large-scale systematic review or evidence synthesis and want to understand not only what thematic clusters exist in a literature but also how they covary with study-level moderators. It is especially valuable when the corpus is heterogeneous in design quality, publication period, or geographic context and you need to know whether certain themes are artifacts of methodological variation. Minimum corpus size is typically 100–200 documents; the method loses statistical stability with smaller corpora. Do not use it when you simply want a descriptive keyword map without moderator hypotheses, or when study-level moderator data are unavailable — in those cases standard co-word analysis or bibliometric coupling is more appropriate.
Strengths & limitations
- Integrates two complementary perspectives — qualitative thematic mapping and quantitative statistical modelling — into a single analytical framework.
- Makes heterogeneity in the literature visible at the thematic level, showing which research clusters are robust across moderator conditions and which are condition-dependent.
- Supports evidence-informed prioritisation of future research by highlighting under-investigated or methodologically weak thematic areas.
- Applicable to any research domain where keyword metadata and study-level variables can be extracted from bibliographic databases.
- Provides a richer foundation for systematic review conclusions than either method alone.
- Requires a sufficiently large corpus (typically 100+ documents) to produce stable co-occurrence frequencies and reliable regression estimates simultaneously.
- Dependent on the completeness and consistency of keyword indexing across databases; missing or idiosyncratic author keywords degrade co-occurrence quality.
- The integration of regression weights into co-word networks involves analytical choices (normalisation measure, weighting scheme) that are not yet standardised, reducing comparability across studies.
- Interpretation demands expertise in both bibliometrics and meta-analytic statistics, making the method less accessible to researchers with training in only one tradition.
Frequently asked
How is this different from ordinary co-word analysis?
Standard co-word analysis builds a keyword network from raw or normalised co-occurrence counts and treats every paper as equally informative. Meta-regression-based co-word analysis re-weights that network using regression estimates of moderator effects, so that the thematic map reflects both co-occurrence patterns and the statistical influence of study-level characteristics. The result is a map that is sensitive to moderator variation, not merely to term frequency.
What software can I use?
The analysis typically combines bibliometric tools (VOSviewer, Bibliometrix in R, or SciMAT) for co-occurrence extraction and network construction with statistical packages (metafor, lme4, or Stata) for meta-regression modelling. The integration step — projecting regression weights onto network edges — is usually implemented via custom R or Python scripts, as no single dedicated software package currently covers the full pipeline.
How large does my corpus need to be?
As a practical threshold, you need enough documents to produce reliable co-occurrence frequencies (generally 100–200 documents minimum) and enough study-level observations to estimate regression coefficients with acceptable precision (at least 20–30 per moderator variable being tested). Smaller corpora can support exploratory applications but should be interpreted with caution.
Can this method replace a standard systematic review?
No. Meta-regression-based co-word analysis is a secondary synthesis technique that describes and maps the structure of a literature. It does not replace the narrative, risk-of-bias assessment, and effect-size synthesis steps of a full systematic review; instead, it complements them by providing a structured picture of how thematic content and study quality are distributed across the corpus.
What normalisation measure should I use for the co-occurrence matrix?
The equivalence index (e_ij = c_ij^2 / (c_i * c_j)) and cosine similarity are the most common choices in co-word analysis. The inclusion index is useful when you want to preserve asymmetric relationships between a niche term and a broad one. For meta-regression-based co-word analysis, the choice of normalisation can interact with the regression weighting scheme, so sensitivity analyses comparing two measures are recommended.
Sources
- Callon, M., Courtial, J. P., Turner, W. A., & Bauin, S. (1983). From translations to problematic networks: An introduction to co-word analysis. Social Science Information, 22(2), 191–235. DOI: 10.1177/053901883022002003 ↗
- Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. Journal of Statistical Software, 36(3), 1–48. DOI: 10.18637/jss.v036.i03 ↗
How to cite this page
ScholarGate. (2026, June 3). Meta-Regression-Based Co-Word Analysis. ScholarGate. https://scholargate.app/en/scientometrics/meta-regression-based-co-word-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bibliometric AnalysisScientometrics↔ compare
- Co-word AnalysisScientometrics↔ compare
- Meta-RegressionMeta Analysis↔ compare
- Science MappingBibliometrics↔ compare