if Tau-squared is 0, do we get any prediction interval? when I run a check on the software, it just shows that studies share a common effect size.. does it mean that there will be no PI available? thanks for all the help
Yes. The first thing is you get a clear sense of how widely the effects vary. If I say that Tau for a risk ratio is 0.02 you don't know what that means. But if I say the effects vary from 0.30 to 0.90 you get the same information on a useful scale. Second, the PI is based on the mean as well as the dispersion. So, you know if the effects vary from 0.30 to 0.90 (helpful in all cases) or 0.70 to 1.3 (helpful in some, harmful in others). Basically, the PI is simply taking tau plus the mean and putting into a metric that people can understand. Sign up at www.Meta-Analysis-Workshops.com to be alerted to related videos. Also www.Meta-Analysis-Books.com has a free PDF explaining this in great detail, with examples.
