{smcl}
{* Mar 21 2010}{...}
{hline}
help for {hi:allsubsets}
{hline}
{title:All subsets (combinatorial) meta-analysis}
{p 4 17 2}
{cmdab: allsubsets} varlist [{hi:if} expression] [{hi:in} range], {hi:s}ave({it:filename})
[{hi:maxstudies}({it:integer}) {hi:replace} ]
{title:Description}
{p 4 8 2}
{cmd:allsubsets} performs all subsets (combinatorial) meta-analysis in a set of studies and saves results.
By default the routine performs fixed effects meta-analysis using the inverse variance method, and random effects meta-analysis using the DerSimonian and Laird random effects model.
Let N be the number of studies in the meta-analysis.
Then the number of possible subsets is -1+2^N.
The command handles up to N={hi:maxstudies} studies. See Options.
{title:Suggested citation:}
{p 4 8 2}
Olkin I, Dahabreh IJ, Trikalinos TA. {browse "http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1053/abstract":GOSH - A graphical display of study heterogeneity}. {it:Research Synthesis Methods}. 2012;3(3):214-223.
See also the {browse "http://www.cebm.brown.edu/resources":Center for Evidence-based Medicine (CEBM)} at Brown University for updates or additional materials.
{hline}
{p 4 8 2}
{hi:varlist} Expects two numeric variables: the effect size {it:Theta} and its variance {it:var_theta}.
{hline}
{p 4 8 2}
In the saved file one finds the following variables:
{p 4 8 2}
{it:study_ES}: the effect size of the study. has non-missing values only for N (# of studies) observations.
{p 4 8 2}
{it:study_var}: the variance of each study. has non-missing values only for N (# of studies) observations.
{p 4 8 2}
{it:es_fem}: fixed effects meta-analysis result
{p 4 8 2}
{it:es_rem}: random effects meta-analysis result
{p 4 8 2}
{it:df}: degrees of freedom
{p 4 8 2}
{it:I2}: I^2 value
{p 4 8 2}
{it:subset}: codes which studies are in the subset. Studies are named with lowercase letters (e.g. "a", "b"). "a b g" means that the 2nd, 3rd and 7th study are included in the subset.
{p 4 8 2}
{it:pRem_z}: p-value for the overall effect with the random effects model
{p 4 8 2}
{it:var_fem}: variance of the summary fixed effects estimate
{p 4 8 2}
{it:var_rem}: variance of the summary random effects estimate
{p 4 8 2}
{it:Q}: Cochran's Q
{p 4 8 2}
{it:tau2}: DerSimonian and Laird 1-step tau^2
{p 4 8 2}
{it:pFem_z}: p-value for the overall effect with fixed effects
{hline}
{p 4 8 2}
Options for {hi:allsubsets}
{p 4 8 2}
{hi:save} Give the nane of the file where the command will save results
{p 4 8 2}
{hi:replace} Whether or not to replace the file if it exists.
{p 4 8 2}
{hi:maxstudies} The maximum number of studies that are allowed. This depends on
your computer's RAM. The default is N=20 (1'048'575 subsets). Specifying too
high an N results in a memory allocation error.