Skip to content

[R-meta] A dataset with 0 heterogeneity

3 messages · Iago Giné Vázquez, Guido Schwarzer

#
Iago,

The PNG file is very hard to read. Could you upload a PDF file?

Not sure what you mean by "second option feasible". Could you please elaborate.

Best,
Guido
#
Guido,

I share the pdf file here: https://ann.nl.tab.digital/s/aNPBFedC6s8eXMR

Regarding the "second option feasible", the second option is the second point you mentioned in your first mail:

- all differences between individual study results can be explained by chance, i.e., random deviations due to (limited) sample sizes in individual studies.


Could it explain the 0 heterogeneity in my example?

Best,
Iago
11 days later
#
Overall, 21 of 22 confidence intervals include the meta-analysis estimate. This is exactly what we would expect under homogeneity for 95% confidence intervals, i.e., 19 of 20 CIs include the meta-analysis estimate. Accordingly, the estimates of tau2 and I2 are 0 which means that all observed differences between point estimates can be explained by chance / random deviations.

Best,
Guido