.-
help for ^metareg^                                            (STB-42: sbe23)
.-

Meta-analysis regression
------------------------

        ^metareg^ y varlist [^if^ exp] [^in^ range], 
        {^wss^e^(^varname^)^ | ^wsv^ar^(^varname^)^ | ^wss^e^(^varname^)^ ^wsv^ar^(^varname^)^}
        [^bs^est^(^{^reml^ | ^ml^ | ^eb^ | ^mm^}^)^ ^tol^eran^(^#^)^ ^l^evel^(^#^)^ ^noit^er]
	

Description
-----------

^metareg^ extends a random effects meta-analysis to estimate the extent
to which one or more covariates, with values defined for each study in
the analysis, explain heterogeneity in the treatment effects.  Examples
of such study-level covariates might be average duration of follow-up,
some measure of study quality, or a measure of geographical location of
each study.  ^metareg^ fits models with two additive components of variance,
one representing the variance within units, the other the variance between
units, and therefore is applicable both to the meta-analysis situation,
where each unit is one study, and to other situations such as multi-center
trials, where each unit is one center. 


Options
-------

^wsse(^varname^)^ is a variable containing the estimated standard
error of y within each study.  Either this or the ^wsvar^ option below 
(or both) must be included.  

^wsvar(^varname^)^ is a variable containing the estimated variance
of y within each study.   Either this or the ^wsse^ option above 
(or both) must be included.  

^bsest(^{^reml^ | ^ml^ | ^eb^ | ^mm^}^)^ specifies the method of estimating the additive 
between studies variance tau^^2.  The default is ^reml^ (restricted maximum 
likelihood using an iterative procedure), with
alternatives being ^ml^ (maximum likelihood), ^eb^ (empirical
Bayes using iterative procedures), or ^mm^ (moment estimator, 
non-iterative).

^toleran(^#^)^ specifies the difference between values of tau^^2 at successive 
iterations required for convergence.  If # is n, the process will not converge
until successive values of tau^^2 differ by less than 10^^-n.
The default is 4.

^level(^#^)^ specifies the confidence level, in percent, for confidence 
intervals.  The default is ^level(95)^ or as set by ^set level^.

^noiter^ requests the log of the iterations be suppressed from the output. 


Remarks
-------

The regression model relates the treatment effect to the study-level
covariates, assuming a normal distribution for the residual errors
with both a within-study and an additive between-studies component
of variance, denoted tau^^2. The within-study variance or standard
error must be supplied by the user. tau^^2 is estimated either by an
iterative procedure, using an estimate which is based on one of restricted 
maximum likelihood (default), maximum likelihood or empirical Bayes
methods, or by a non-iterative procedure using a method of moments estimator.

The estimated between-studies variance tau^^2 is a measure of the residual
hetereogeneity having adjusted for the covariates.  If no covariates are
specified, the regression is equivalent to a random effects meta-analysis;
note: if the mm option is chosen for estimation of tau^^2, the estimated
value should correspond to that obtained from the command @meta@.


Examples
--------

   . ^metareg logor , wsse(selogor) bs(mm)^  
  /* results equivalent to ^meta logor selogor^ */

   . ^metareg logor dur, wsvar(vlor) bs(eb) noit^
   . ^metareg meandiff qual avchol, wsse(sediff) bs(ml) tol(5) l(90)^ 


Author
------

	Stephen Sharp
	London School of Hygiene and Tropical Medicine
	e-mail: stephen.sharp@@lshtm.ac.uk


Reference
---------

Thompson, S. G. and S. J. Sharp. 1998. Explaining heterogeneity in 
meta-analysis: a comparison of methods. submitted.


Also see
--------

    STB:  STB-42 sbe23
On-line:  help for @meta@ (if installed), @metacum@ (if installed), 
          @metabias@ (if installed)