3

I want to ask whether I can calculate the Cochrane $Q$ and $I^2$ statistics in a random effect model.

I collected some data from different areas and want to compare the prevalence of allergic sensitization on different areas.

However, the Cochrane $Q$ and $I^2$ statistics were quite large (usually, >= 90 for $I^2$) and I can't describe the heterogeneity of prevalence of allergen sensitization between areas.

So, if possible, I want to calculate the $Q$ and $I^2$ statistics by using new weight w'

$w'$= 1/(Var($E$)+$T^2$)

$T^2$ = $Q$-($k$-1)/(Σ$w$-Σ$w^2$/Σ$w$)

$w$=1/Var($E$)

However, I'm not sure that such statistics could be applied in a random effect model.

mdewey
  • 16,541
  • 22
  • 30
  • 57
Mingyu Kang
  • 31
  • 2

1 Answers1

1

Michael Borenstein has a worked example in Excel (http://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CCkQFjAA&url=http%3A%2F%2Fwww.meta-analysis.com%2Fdownloads%2FIntroduction%2520to%2520Meta-Analysis%2520Data%2520files%2520Chapter%252019.xls&ei=a1anVPGUBIiEyQSb_YCQAw&usg=AFQjCNHT1IfDnCnNSTW3u_k2F_w0XAXieQ&sig2=1PIWW0u9xrBhwo4A4pDpeA&bvm=bv.82001339,d.aWw&cad=rja). He describes this in detail in his book 'Introduction to Meta-Analysis', Chapter 19.

abousetta
  • 1,030
  • 5
  • 9
  • For the benefit of those of us without access to the book is it possible for you to say what interpretation Borenstein puts on Q and I^2? – mdewey Feb 29 '16 at 15:52