It happens that the results with p>0,05 in primary studies are not reported or there is only a short note "the results for (...) were statistically not significant". However, in situation where authors reported the results with adnotation p>0,05 should these results be included in meta-analytic study? By doing so (adding the "non-significant" results to meta-analysis), I believe, we could avoid bias connected with reporting only the "significant" results. I would be extremely grateful for an advice in this matter.
You are completely correct.
A p-value is got by dividing the effect size by its standard error, which measure the precision with which the effect size was measured. So the same p-value can result from a trivial effect size measured with great precision or an impressive effect size measured with low precision. So "significant" studies are a strange mixture of studies with no necessary (or even definable) applicability.
Dear Professor
Thank you for your answer. There is an ongoing discussion nowadays regarding the p-value and your important answer perfectly decribes the point of it. Thank you for your precious time.
Hi, in a SR it is important to differentiate biases induced by the reviewer of biases in individual studies. I am aware that your question is about MA, but MA as a part of a SR are more likely to be of high methodological quality.
First you assess risk of bias in individual studies.
The fact that their authors do not report non-statistically significant results is of concern about risk of bias by selective reporting (you Also should be aware if they are performing multiple testing looking for p
you include both in the presentation so that you can better make a case.
I think, If meta to be done on the results of analytical studies, should all effects entered the meta, significant or non-significant
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size