I am conducting a meta-analysis of studies where both the exposure and outcome variables are continuous. The studies typically report a combination of standardised and unstandardised beta coefficients, with varying availability of accompanying standard errors. I am struggling to find beginner-level advice on how to combine these sorts of effect estimates in meta-analysis (and indeed whether I should combine them at all!)
Any signposting to helpful guides on this would be much appreciated.
It is unclear as to what you are trying to do. State your research question. It sort of sounds like you are trying to do a meta-analysis on studies reporting correlations. The best beginner-level meta-analysis book in my opinion is Introduction to Meta-Analysis by Borenstein, Hedges, Higgins, and Rothstein. In Chapter 6, it addresses using correlations as the effect size in a meta-analysis.
It sounds to me like you have beta coefficients for dose related risk from regrssions in the individual studies. Chapter 32 from Rothman and Greenland Modern Epidemiology 2nd ed 1998 (p 643-673) deals well with this including dealing with missing SEs or CIs as long as you have denominator sizes in the individual studies.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis