Title
metaeff -- Meta-analysis module for effect sizes calculations
Syntax
metaeff varname1 varname2 [if] [in] [, options]
where
varvame1 the variable corresponding to the estimated effect sizes.
varvame2 the variable corresponding to the estimated standard errors of the effect sizes.
options Description ------------------------------------------------------------------------- Options ni(varname) specifies the number of units in the intervention group nc(varname) specifies the number of units in the control group i(varname) specifies the number of events in the intervention group c(varname) specifies the number of events in the control} group orval(varname) specifies the odds ratio ci95lo(varname) specifies the lower end of 95% confidence interval of the mean difference or odds ratio ci95up(varname) specifies the upper end of 95% confidence interval of the mean difference or odds ratio md(varname) specifies the mean difference (not standardised) meani(varname) specifies the mean of the intervention group meanc(varname) specifies the mean of the control group sdi(varname) specifies the standard deviation in the intervention group sdc(varname) specifies the standard deviation in the control group ci95loi(varname) specifies the lower end of the 95% confidence interval of the mean for the intervention group ci95upi(varname) specifies the upper end of the 95% confidence interval of the mean for the intervention group ci95loc(varname) specifies the lower end of the 95% confidence interval of the mean for the control group ci95upc(varname) specifies the upper end of the 95% confidence interval of the mean for the control group p(varname) specifies the p value for a two-sample t-test t(varname) specifies the t value for a two-sample t-test (assumes df=ni+nc-2) or selects odds ratio (1b) over the default risk difference method (1a), when ni, nc, i & c are available owrite specifies that varvame1 and varvame2 will be overwritten, if they already exist in the dataset infovar provides information on the method selected (for each case) in a new string variable _method
Description
The metaeff command provides a way to calculate the effect sizes (and the respective standard errors) of research studies, for use with meta-analysis methods (see meta, metan & metaan). The methods used for the calculations have been derived from the Cochrane Collaboration handbook. The calculated effect sizes are standardised mean differences and binary data are transformed towards that end. metaeff is meant to be used for meta-analysing studies with various types of outcomes; if all studies provide binary outcomes then metaeff should not be used since better approaches exist and a transformation is not required.
Dichotomous data
Method 1a, based on Risk Difference. Requires ni, nc, i & a.
Method 1b, based on Odds Ratio. Requires ni, nc, i & a.
Method 2, based on Odds Ratio. Requires ni, nc, orval, ci95lo & ci95up.
Continuous data
Method 3, based on Mean Difference. Requires ni, nc, md or (meani and meanc) , ci95lo & ci95up.
Method 4, based on Mean Difference. Requires ni, nc, md or (meani and meanc), sdi & sdc.
Method 5, based on Mean Difference. Requires ni, nc, md or (meani and meanc), ci95loi, ci95upi, ci95loc, ci95upc.
Method 6, based on Mean Difference. Requires ni, nc, md or (meani and meanc) & p or t.
Dichotomous or Continuous data
Method 7, based on z-value. Requires ni, nc & p or t.
Remarks
Some methods are more accurate than others and have been prioritised accorgingly (in case user has provided input that allows 2 or more methods to be selected). For dichotomous data the method order is 2, 1a, 1b, 7 and for continuous data 4, 5, 3, 6 & finally 7. Methods 6 and (especially) 7 should only be used if no other method can be selected (due to distibutional assumptions and the use of the reported p-value is usually inaccurate). The methods are described in Kontopantelis and Reeves, 2009 and have been prioritised according to expected precision: that is, the expectation that the effect size and associated variance computed from the input data will be accurate. As a general rule, the fewer the number of mathematical transformations involved in getting from the 'raw data' to the statistical parameters used as input for the method, the higher the expected precision.
Examples
. metaeff eff SEeff, ni(ivsize) nc(clsize) i(ivevents) c(ivevents)
. metaeff eff SEeff, ni(ivsize) nc(clsize) i(ivevents) c(ivevents) or owrite
. metaeff eff2 SEeff2, ni(ivsize) nc(clsize) md(meandiff) sdi(ivsd) sdc(clsd)
. metaeff eff SE, ni(ivsize) nc(clsize) meani(ivmean) meanc(clmean) ci95loi(ivlo) ci95upi(ivup) ci95loc(cllo) ci95upc(clup)
. metaeff eff SEeff, ni(ivsize) nc(clsize) t(tvalue)
Authors
Evangelos Kontopantelis, National Primary Care Research and Development Centre,
University of Manchester, e.kontopantelis@manchester.ac.uk
David Reeves, Health Sciences Primary Care Research Group, University of Manchester
References
Cochrane Collaboration handbook: http://www.cochrane.org/resources/handbook/ Kontopantelis, E. and Reeves D. 2009. A Meta-Analysis add-in for Microsoft Excel. Journal of Statistical Software.
Also see
STB: STB-44 sbe24
help for metaan, metan7 (if installed)
metannt (if installed), meta (if installed)
metacum (if installed), metareg (if installed)
metabias (if installed), metatrim (if installed)
metainf (if installed), galbr (if installed)
metafunnel (if installed)