If the mean difference was used to measure the effect estimates in comparisons, in which threshold value we should perform rating up the quality of evidence in Grade analysis. In terms of Risk ratio, 2< is considered a large effect, but in continuous variables how can we define magnitudes of effects to perform rating up in grade analysis.
I am uncertain if I fully understand your question. If you are looking for benchmarks for effect sizes calculated for continuous variables using means, the only ones I am aware of are for standardized mean difference (aka Cohen's d). 0.2 is considered small, 0.5 is considered moderate, and 0.8 is considered large (Cohen J. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988:567). These benchmarks also work for Hedges' g.
Ömer Hatipoğlu GRADE quality of evidence is not concerned by the effect estimate itself, so I'm not sure what are you actually asking. You might be interested in the precision of that estimate (one of the GRADE criteria), and then you'll have to look at corresponding confidence intervals and make a decision whether you should downgrade level of evidence for one level for that particular criteria in the GRADE assessment. If you were just interested in some "thresholds" to communicate the magnitude of your effect estimates, then just follow the advice from Gordon L Warren although bear in mind that they might not be applicable (i.e. meaningful) to all fields, and placing too much emphasis exclusively on them – regardless of which set of thresholds you pick – is probably not a good idea either.
To my knowledge, there has yet to be a certain threshold for upgrading continuous outcomes for large-magnitude of effect. I would suggest converting MD to SMD first and then SMD to OR/RR. This conversion may help you to use >2 RR threshold for upgrading GRADE.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis