{smcl}
{* *! version 2.01 01jan2022}{...}
{viewerdialog maanova "dialog maanova"}{...}

{marker syntax}{...}

{title:Syntax}

{phang} Subgroup (categorica) fixed- and random-effects
inverse-variance weighted meta-analysis.

{p 8 16 2}
{cmd:maanova} {effectsizevar} {catvar} {ifin} {cmd:,}
{opt var(varname)}
{opt w(varname)}
{opt se(varname)}
[_options_]

{pstd} where {it:effectsize} is the effect size variable and
{it:catvar} is a categorical variable. The {it:effect size} can be any
effect size type, such as Cohen's {it:d}, Hedges' {it:g}, logged odds
ratio, logged risk ratio, logged logit, {it:r} or Fisher's {it:Zr},
among others. It is critical that the effect size is in its analyzable
form. For example, the effect size can be a logged odds ratio but not
an odds ratio. One of the following must also be specified:

{phang2}
o  {it:var({varname})}: the variance of the effect size

{phang2}
o  {it:w({varname})}: the inverse variance weight of the effect size

{phang2}
o  {it:se({varname})}: the standard error of the effect size

{pstd} The relationship among these is assumed to be {it:w} =
1/{it:var} = 1/{it:se^2}.  {p_end}

{marker reoptions}{...}
{synoptset 20 tabbed}{...}
{synopthdr :Options}
{synoptline}
{syntab:Model Type}
{synopt :{opt model(_string_)}}
 model type; default is REML (restricted maximum likelihood); options
 include FE (fixed effect), DL (Dersimonian & Laird), HE (Hedges'), HS
 (Hunter & Schmidt), SJ (Sidik-Jonkman), SJIT (Sidik-Jonkman,
 iterative), ML (maximum likelihood), REML (restricted maximum
 likelihood), and EB (empirical bayes)
{p_end}

{syntab:Print Options}
{synopt :{opt print(_string_)}}
print options convert results for ease of interpretation; {it:exp}
exponentiates results, {it:ivzr} is the inverse Fisher's {it:z} transformation,
producing {it:r}, and {it:prop} converts logits back into proportions
{p_end}

{syntab:Tau^2 Options} {synopt :{opt tau_unique(_string_)}} specifies
whether a common (default) or subgroup specific tau^2 is
used. {it:tau_unique(YES)} will estimate a separate tau^2 for each
subgroup.  {p_end}

{marker description}{...}
{title:Description}

{pstd} {cmd:maanova} performs a subgroup or categorical moderator
meta-analysis under either a fixed-effect model (also called a
common-effect model) or a random-effects model. Several estimators for
the random effects variance component (tau^2) are available.  The
command requires an effect size, its associated standard error,
variance, or inverse variance weight, and a categorical variable.

{pstd} Meta-analytic regresson (aka, meta-regression) can be performed
with the {cmd:mareg} command (see {help mareg}). For a basic
meta-analysis returning the overall mean effect size and associated
statistics, see {cmd:masum} command (see {help masum}).

{pstd} As of Stata version 16.0, Stata has a built-in command for
conducting meta-analysis. See {help meta}.

{marker description}{...}
{title:Acknowledgments}

{pstd} {cmd:maanova} was written by David B. Wilson and is an updated
version of a command written as a companion to a book on meta-analysis
he co-authored with Mark Lipsey (Lipsey & Wilson, 2001). Portions of
this program are based on code from Wolfgang Viechtbauer's
{it:metafor} package for R.

{marker description}{...}
{title:References}

{pstd}
Lipsey, M. W., & Wilson, D. B. (2001). {it} Practical
meta-analysis. {sf} Sage.