[R] Meta-Analysis of proportions

[Chuck Cleland]

Monica Malta wrote:
> Dear colleagues,
> 
> I'm conducting a meta-analysis of studies evaluating adherence of HIV-positive drug users into AIDS treatment, therefore I'm looking for some advice and syntax suggestion for running the meta-regression using proportions, not the usual OR/RR frequently used on RCT studies.
> 
> Have already searched already several handbooks, R-manuals, mailing lists, professors, but... not clue at all...
> 
> Does anyone have already tried this? A colleague of mine recently published a similar study on JAMA, but he used OpenBUGS - a software I'm not familiar with...
> 
> If there is any tip/suggestion for a possible syntax, could someone send me? I need to finish this paper before my PhD qualify, but I'm completely stuck...
> 
> So, any tip will be more than welcome...I will really appreciate it!!! 
> 
> Thanks in advance and congrats on the amazing mailing-list.

  Specifying adherence as the "effect size", if you can also specify a
sampling variance perhaps you can use a tool like mima
(http://www.wvbauer.com/downloads/mima_tutorial.pdf).  For example,
working with logits rather than proportions directly, consider the
following example data:

> df
   ntot nadhere mod      prop      logit        se
1   100      76   0 0.7600000 -1.1526795 0.2341463
2   125      98   0 0.7840000 -1.2891306 0.2173501
3    50      37   0 0.7400000 -1.0459686 0.3224127
4   200     159   0 0.7950000 -1.3553321 0.1751557
5   150     114   0 0.7600000 -1.1526795 0.1911796
6    80      56   1 0.7000000 -0.8472979 0.2439749
7   160     113   1 0.7062500 -0.8772402 0.1735688
8   200     130   1 0.6500000 -0.6190392 0.1482498
9   120      75   1 0.6250000 -0.5108256 0.1885618
10  105      78   1 0.7428571 -1.0608720 0.2232879

  Then do the meta-regression as follows:

> source("http://www.wvbauer.com/downloads/mima.ssc")

> with(df, mima(yi=logit, vi=se, mods=mod))

Estimate of (Residual) Heterogeneity: 0

Test for (Residual) Heterogeneity:

QE      =  1.2419
df      =  8
p-value =  0.9962

Parameter Estimates:

           [,1]
intrcpt -1.2161
mods     0.4520

Variance-Covariance Matrix of Parameter Estimates:

        intrcpt    mods
intrcpt  0.0436 -0.0436
mods    -0.0436  0.0815

Omnibus Test of all Moderators:

QME     =  2.5058
df      =  1
p-value =  0.1134

Individual Moderator Tests:

        estimate     SE    zval   pval    CI_L    CI_U
intrcpt  -1.2161 0.2089 -5.8213 0.0000 -1.6256 -0.8067
mods      0.4520 0.2856  1.5830 0.1134 -0.1077  1.0117

> Bests from Rio de Janeiro, Brazil.
> 
> Monica  
> 
> Monica Malta
> Researcher
> Oswaldo Cruz Foundation - FIOCRUZ
> Social Science Department - DCS/ENSP
> Rua Leopoldo Bulhoes, 1480 - room 905
> Manguinhos
> Rio de Janeiro - RJ 21041-210
> Brazil
> phone +55.21.2598-2715
> fax +55.21.2598-2779
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> 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.

-- 
Chuck Cleland, Ph.D.
NDRI, Inc.
71 West 23rd Street, 8th floor
New York, NY 10010
tel: (212) 845-4495 (Tu, Th)
tel: (732) 512-0171 (M, W, F)
fax: (917) 438-0894