Different weightings in metan

[Jingjing Niu]

Hi, Statalist

I am currently using the -metan- package for my meta-analysis. Yet, I found the weights assigned to each study were very different for different outcomes (standardized mean difference (SMD) versus weighted/unstandardized mean difference (WMD)). When I was checking around, I found a third kind of weighting which confused my even more. I have two questions regarding the following results:
1. Is it normal to have very different weightings? If not, what is wrong?
2. Which kind of pooled ES is calculated by the "metan var(mean of change) var(se of change)... (i.e. "metan fvdiffm fvdiffse" below)"? The ES for each study is the same with WMD, yet the weights and pooled ES is different from both WMD and SMD. Please check the results below:

	metan fvdiffm fvdiffse		Hedge's g		nostandard
Study	ES	% Weight	SMD	% Weight	WMD	% Weight
a (2010)	0.17	12.9	0.099	21.52	0.17	31.39
b (2010)	0.17	12.9	0.099	21.51	0.17	31.38
c (2009)	-0.71	10.17	-0.404	3.26	-0.71	1.54
d (2008)	0.14	12.83	0.082	19.45	0.14	22.89
e (2014)	1.08	8.94	0.517	2.88	1.08	1.01
f (2014)	2.4	12.6	0.83	12.72	2.4	0.98
g (2011)	0.5	11.69	0.23	9.42	0.5	3.78
h (2008)	1.22	5.73	0.411	2.11	1.22	0.36
i (2004)	-0.1	12.24	-0.079	7.14	-0.1	6.67

Pooled ES: 0.505 0.192 0.179

Code:
gen fvdiffse=fvdiffsd/sqrt(nw+nm)
gen n=nw+nm

metan fvdiffm fvdiffse if code==0, random label(namevar=author, yearvar=year) ///
xtitle(Mean difference in F/V intake change) xlabel(-1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2, 2.5, 3)


metan nw fvchangemw fvchangesdw nm fvchangemm fvchangesdm if code==0, hedges random label(namevar=author, yearvar=year) ///
xtitle(Mean difference in F/V intake change) xlabel(-0.25, -0.20, -0.15, -0.10, -0.05, 0, 0.05, 0.10, 0.15)


metan nw fvchangemw fvchangesdw nm fvchangemm fvchangesdm if code==0, nostandard random label(namevar=author, yearvar=year) ///
xtitle(Mean difference in F/V intake change) xlabel(-1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2, 2.5, 3)


Thank you very very much for your help!
Jingjing