META-SPSS
DISCLAIMER
----------
META-SPSS is provided "as is" without warranty of any kind. The entire risk
as to the quality, performance, and fitness for intended purpose is with you.
You assume responsibility for the selection of the program and for the use of
results obtained from that program.
DESCRIPTION
-----------
This collection of files adds meta-analytic facilities to SPSS:
1. Meta-analysis of P values (see 'MetaPValues-Info.txt' for technical details).
2. Meta-analysis of binary outcomes: RD, OR and RR, with both fixed & random
effects (DerSimonian-Laird) models. Raw data (counts) or summary data
(adjusted OR, RR or RD) can be used as input. Different fixed effects models:
inverse variance (Woolf method), Mantel-Haenszel and Peto (this last only
for OR) are available for raw data (counts); for summary data only the first.
3. Meta-analysis of continuous outcomes: unstandardised & standardised (Hedges,
Cohen & Glass) mean differences, with both fixed (inverse variance) & random
effects (DerSimonian-Laird) methods. Means, SD & N or Cohen's d can be used
for input (except for Glass method, which requires means, SD & N).
Orwin's Fail Safe N is computed for significant results.
4. Meta-analysis of correlation coefficients: Hedges-Olkin fixed & random
effects models (inverse variance & Dersimonian-Laird).
Orwin's Fail Safe N is computed for significant results.
5. Meta-analysis of correlation coefficients: Schmidt-Hunter method (simplified,
unreliability or range departure correction methods are not included). See
Larry Lyons paper & website (details in folder "Documents") for details.
For methods 1 to 4, Higgins' I2 with its 95%CI is calculated
(see 'Higgins I2 formula.txt' for details and a reference).
For methods 2, 3 & 4, forest plots are also included (see below for a limitation).
You must know meta-analysis before using them, although some short documents with
meta-analysis theory & formulae are included in the folder "Documents". I recommend
reading Andy Field's papers (Meta1.pdf, Metar.pdf & Metad.pdf)and taking a look at
his website. A text file with useful references (mainly from British Medical Journal)
and tutorials is also included.
A very good book on the topic is 'Systematic Reviews in Health Care' Edited by
Mathias Egger, George Davey Smith & Douglas G Altman. BMJ Books, 2nd Ed (2001).
(ISBN: 0-7279-1488-X).
Feel free to edit & modify every file. If you improve the code, I would like to
have a copy.
Questions can be sent to me: Marta Garcia-Granero (biostatistics@terra.es).
SPSS files included:
-------------------
- MetaDatasets.SPS: creates in drive A: a collection of datasets from published
meta-analysis, mainly from British Medical Journal. Must be run once to obtain
the datasets that MetaTests.SPS requires.
- MetaAnalysisOfxxxxxxx.SPS: Syntax files for meta-analytic methods 1 to 5.
The name, a bit long, it's related to the method:
Method File
----------------------------------------------------------
Odds Ratio:
Peto MetaAnalysisOfBinOutcomes(OR-Peto).SPS
M-H MetaAnalysisOfBinOutcomes(OR-MH).SPS
Inv Var MetaAnalysisOfBinOutcomes(OR-IV).SPS
DerS-Laird MetaAnalysisOfBinOutcomes(OR-DL).SPS
Summary data MetaAnalysisOfBinOutcomes(SumOR-IV).SPS
MetaAnalysisOfBinOutcomes(SumOR-DL).SPS
Risk Ratio:
M-H MetaAnalysisOfBinOutcomes(RR-MH).SPS
Inv Var MetaAnalysisOfBinOutcomes(RR-IV).SPS
DerS-Laird MetaAnalysisOfBinOutcomes(RR-DL).SPS
Summary data MetaAnalysisOfBinOutcomes(SumRR-IV).SPS
MetaAnalysisOfBinOutcomes(SumRR-DL).SPS
Risk Diff:
M-H MetaAnalysisOfBinOutcomes(RD-MH).SPS
Inv Var MetaAnalysisOfBinOutcomes(RD-IV).SPS
DerS-Laird MetaAnalysisOfBinOutcomes(RD-DL).SPS
Summary data MetaAnalysisOfBinOutcomes(SumRD-IV).SPS
MetaAnalysisOfBinOutcomes(SumRD-DL).SPS
Mean Difference:
Inv Var MetaAnalysisOfContOutcomes(WM-IV).SPS
DerS-Laird MetaAnalysisOfContOutcomes(WM-DL).SPS
Hedges G (Stand. Mean Difference):
Inv Var MetaAnalysisOfContOutcomes(G-IV).SPS
DerS-Laird MetaAnalysisOfContOutcomes(G-DL).SPS
Summary data MetaAnalysisOfContOutcomes(SumG-IV).SPS
MetaAnalysisOfContOutcomes(SumG-DL).SPS
Cohen d (Stand. Mean Difference):
Inv Var MetaAnalysisOfContOutcomes(D-IV).SPS
DerS-Laird MetaAnalysisOfContOutcomes(D-DL).SPS
Summary data MetaAnalysisOfContOutcomes(SumD-IV).SPS
MetaAnalysisOfContOutcomes(SumD-DL).SPS
Glass delta (Stand. Mean Difference):
Inv Var MetaAnalysisOfContOutcomes(Delta-IV).SPS
DerS-Laird MetaAnalysisOfContOutcomes(Delta-DL).SPS
Correlation (Hedges-Olkin):
Inv Var MetaAnalysisOfCorrCoef(HO-IV).SPS
DerS-Laird MetaAnalysisOfCorrCoef(HO-DL).SPS
Correlation (Schmidt-Hunter):
MetaAnalysisOfCorrCoef(SH).SPS
P Values:
MetaAnalysisOfPvalues.SPS
----------------------------------------------------------
- ForOR.SPS: For binary outcomes the information included assumes that RR
and RD are the summary data of interest. This syntax has to be run after
loading binary outcomes files if OR is the summary measure of interest.
ForOR.SPS creates 4 extra variables, called: caex, canex, coex & conex.
- MetaTests.SPS: a syntax file that shows the use of the different method
files.
- XXXXXX.sct: several chart templates that work with SPSS 11.5 or older.
- Two SAV files in "Temp" folder: they are created by files of methods 2-4.
They were originally included in the ZIP file to be sure that the "Temp"
folder is created during installation, because several syntax files store
information in Temp and it must exist to avoid errors.
INSTALLATION
------------
- Download META-SPSS.ZIP and unzip its contents to a floppy disk (drive A:).
- Run MetaDatasets.SPS once to create the 12 example datasets.
- Open MetaTests.SPS and run it to see how the files work.
ACCURACY OF THE METHODS
-----------------------
I have tried to check the output with published datasets and other programs
(Review Manager, Meta, Easy Meta Analysis, Combine...) and, to my knowledge,
everything is OK.
An extra check was performed comparing META-SPSS output with tables 15.4 & 15.6
of 'Systematic Reviews in Health Care' (p 303-306), using the same datasets.
If you want to check that results with META-SPSS are correct, the datasets are
included as 'MetaDataset(XII).sav' (binary outcomes) & 'MetaDataset(XIII).sav'
(continuous outcomes).
Table 15.4
Method Estimate (95% CI) Sig of effect(*) Heterogeneity
---------------------------------------------------------------------------
Odds Ratio:
Peto 0.59 (0.51 to 0.69) z=7.05 (p<0.001) X2=38.5 (p<0.001)
M-H 0.59 (0.51 to 0.69) z=6.98 (p<0.001) X2=38.9 (p<0.001)
Inv Var 0.60 (0.52 to 0.70) z=6.70 (p<0.001) X2=38.8 (p<0.001)
DerS-Laird 0.54 (0.34 to 0.85) z=2.64 (p=0.008)
Risk Ratio:
M-H 0.79 (0.74 to 0.85) z=6.95 (p<0.001) X2=29.8 (p<0.001)
Inv Var 0.80 (0.75 to 0.85) z=7.14 (p<0.001) X2=29.7 (p<0.001)
DerS-Laird 0.77 (0.64 to 0.92) z=2.93 (p=0.003)
Risk Diff:
M-H -0.117 (-0.149 to -0.085) z=7.13 (p<0.001) X2=33.1 (p<0.001)
Inv Var -0.127 (-0.158 to -0.095) z=7.86 (p<0.001) X2=32.7 (p<0.001)
DerS-Laird -0.124 (-0.211 to -0.038) z=2.81 (p=0.005)
---------------------------------------------------------------------------
Table 15.6
Method Estimate (95% CI) Sig of effect(*) Heterogeneity
---------------------------------------------------------------------------
Fixed effect models:
Cohen's d -0.16 (-0.41 to 0.08) z=1.29 (p=0.20) X2=2.34 (p=0.31)
Hedges' g -0.16 (-0.41 to 0.08) z=1.29 (p=0.20) X2=2.31 (p=0.32)
Glass's d -0.16 (-0.40 to 0.09) z=1.24 (p=0.21) X2=2.28 (p=0.32)
Random effects models:
Cohen's d -0.15 (-0.42 to 0.12) z=1.12 (p=0.26)
Hedges' g -0.15 (-0.42 to 0.11) z=1.12 (p=0.26)
Glass's d -0.15 (-0.42 to 0.12) z=1.10 (p=0.27)
---------------------------------------------------------------------------
(*) META-SPSS gives a X2 test, mathematically equivalent (X2=(Z)2).
---------------------------------------------------------------------------
LIMITATIONS & KNOWN BUGS
------------------------
Cumulative meta-analysis is not possible... yet (I'm working on that). Check
this website from time to time to see if I have updated META-SPSS.
The syntax files assume that everything runs from a floppy disk (A:). You will
have to edit every 'Meta...SPS' file (including MetaTests.SPS & MetaDatasets.SPS)
and replace every "A:" by a suitable drive&path if you want to run them from
other location.
This code was developed before SPSS 12 was released, and the chart templates
included will only work with SPSS 11.5 or older. The graphs will need formatting
if SPSS 12 is used (good luck, I never succeeded trying to apply chart templates
in v12, and some options, available in older versions, have dissapeared).