Dear Mario
I do not use metasens myself so cannot be of direct help but I have
looked at your dataset and it does seem rather strange (as you perhaps
know). You have two quite large studies with very large hazard ratios
and if we ignore them all the rest of the studies fall on a diagonal
bacn indicative of extreme small study bias.
One thing you could consider is to use metafor and within it use the hc
function which uses a different approach due to Henmi and Copas (the
same Copas).
On 12/08/2015 15:19, petre...@unina.it wrote:
Dear all,
I use R 3.1.1 for Windows (x 64).
I performed a meta-analysis of hazard ratio using the below reported
Dataset and metagen function from package meta.
meta1<-metagen(Dataset$lnHR, Dataset$seHR, sm="HR")
Thereafter, I try to use the copas function from package metasens.
cop1<-copas(meta1)
and I have these 3 warnings:
Warning in sqrt(solve(junk2$hessian + 1e-08)[1, 1]) :
NaN was produced
Warning in sqrt(solve(junk2$hessian + 1e-08)[1, 1]) :
NaN was produced
Warning in sqrt(solve(junk2$hessian + 1e-08)[1, 1]) :
NaN was produced
If I try:
plot (cop1)
I have:
ERROR:
object "is.relative.effect" not found
Any suggestion is welcome.
The Dataset is:
id Year lnHR seHR
1 1 2001 0.6881346 0.06940859
2 2 2001 1.4036430 0.60414338
3 3 2002 0.7419373 0.28897730
4 4 2003 1.5475625 0.45206678
5 5 2003 1.4816046 0.44859666
6 6 2005 0.9162908 0.17166950
7 7 2006 1.2697605 0.34205049
8 8 2009 0.8960880 0.24626434
9 9 2011 1.5040774 0.24683516
10 10 2012 0.4510756 0.17213355
11 11 2008 0.9895412 0.26590857
12 12 2009 2.8094027 0.61304092
13 13 2010 0.9162908 0.21362771
14 14 2011 0.5068176 0.15060408
15 15 2012 3.0027080 0.27239493
16 16 2013 1.9837563 0.55793673
17 17 2013 3.0492730 0.18798657
18 18 2014 1.2974632 0.44759619
19 19 2014 0.8241754 0.39551640
20 20 2014 2.2617631 0.56545281
The code used are:
meta1<-metagen(Dataset$lnHR, Dataset$seHR, sm="HR")
meta1
HR 95%-CI %W(fixed) %W(random)
1 1.99 [ 1.7369; 2.2800] 42.92 5.99
2 4.07 [ 1.2455; 13.2997] 0.57 3.71
3 2.10 [ 1.1919; 3.7000] 2.48 5.28
4 4.70 [ 1.9378; 11.3998] 1.01 4.47
5 4.40 [ 1.8264; 10.5998] 1.03 4.49
6 2.50 [ 1.7857; 3.5000] 7.02 5.75
7 3.56 [ 1.8209; 6.9599] 1.77 5.03
8 2.45 [ 1.5120; 3.9700] 3.41 5.47
9 4.50 [ 2.7740; 7.2999] 3.39 5.47
10 1.57 [ 1.1204; 2.2000] 6.98 5.75
11 2.69 [ 1.5974; 4.5300] 2.92 5.38
12 16.60 [ 4.9921; 55.1988] 0.55 3.67
13 2.50 [ 1.6447; 3.8000] 4.53 5.60
14 1.66 [ 1.2357; 2.2300] 9.12 5.81
15 20.14 [11.8085; 34.3497] 2.79 5.36
16 7.27 [ 2.4357; 21.6996] 0.66 3.94
17 21.10 [14.5971; 30.4998] 5.85 5.69
18 3.66 [ 1.5223; 8.7999] 1.03 4.49
19 2.28 [ 1.0502; 4.9499] 1.32 4.76
20 9.60 [ 3.1693; 29.0794] 0.65 3.90
Number of studies combined: k=20
HR 95%-CI z p.value
Fixed effect model 2.7148 [2.4833; 2.9679] 21.9628 < 0.0001
Random effects model 3.9637 [2.7444; 5.7247] 7.3426 < 0.0001
Quantifying heterogeneity:
tau^2 = 0.5826; H = 3.56 [3.04; 4.16]; I^2 = 92.1% [89.2%; 94.2%]
Test of heterogeneity:
Q d.f. p.value
240.64 19 < 0.0001
Details on meta-analytical method:
- Inverse variance method
- DerSimonian-Laird estimator for tau^2
cop1<-copas(meta1)
Warning in sqrt(solve(junk2$hessian + 1e-08)[1, 1]) :
NaN was produced
plot (cop1)
ERROR:
object "is.relative.effect" not found
-------------------------------------------------------
Mario Petretta
Associate Professor of Internal Medicine
Department of Translational Medical Sciences
Naples University Federico II Italy
----
5x1000 AI GIOVANI RICERCATORI
DELL'UNIVERSITÀ DI NAPOLI
Codice Fiscale: 00876220633
www.unina.it/Vademecum5permille
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
--
Michael
http://www.dewey.myzen.co.uk/home.html
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
