MIX 2.0 Lite and Pro

MIX 2.0 Lite and Pro are very much alike when it comes to analytical functions. The big difference is the data that you can analyze. MIX Lite is meant for learning and teaching and therefore only allows the user to analyze built-in datasets. It also does not include some of the extra modules that are developed for MIX 2.0 Pro.

Feature
Lite
Pro
O
O
X
O
Data convertor
X
C
Basic numerical meta-analysis
O
O
Basic graphical meta-analysis
O
O
Meta-regression
X
C
Sample size calculations
X
C
R & OpenBUGS code creators
X
C

O = available, X = unavailable, C = under construction

MIX 2.0 general features

Load built-in datasets (Lite & Pro)

MIX Lite and Pro both contain a large number (currently 20+) built-in data sets. The data sets are from authoritative books on meta-analysis and will be expanded with data sets from methodological articles. MIX can therefore be used to reproduce and expand on the analyses and examples presented in these books. The dataset file is regularly updated and will be available for free download from the MIX 2.0 website.

Create custom datasets (Pro)

MIX 2.0 Pro contains a datawizard that provides a number of options to create custom datasets. The datasets can be of three types:

  • dichotomous outcome data (events and non-events)
  • continuous outcome data (means, standard deviations and group sizes)
  • generic data (study estimate with standard error)


We are currently experimenting with creating another separate option for correlation data or using the data convertor to transform such data to the generic data type described above. 

The data can be entered in a dataset template, either manually or via copy-paste procedures or they can be imported in spreadsheet format. In the latter case, MIX 2.0 Pro will create the dataset after the user has identified the rows that contains the data. MIX 2.0 Pro will soon be able to import datasets that were created in RevMan 4 and RevMan 5.

Data converter (Pro - under construction)

The Data converter in MIX 2.0 Pro enables users to convert data types and variability measures. If a study provides only a correlation coefficient and a sample size, the data can currently not be entered directly in MIX 2.0. However, the converter can for example calculate Fisher's Z and its standard error and make the data compatible with the MIX data entry options. The converter can also be useful if a study only provides an estimate with a confidence interval, but you really want the estimate and its standard error. The convertor can use data of a single study at a time or apply the conversion to a number of studies at once.

Basic numerical meta-analysis (Lite & Pro)

The basic numerical meta-analysis possibilities in MIX 2.0 Lite and Pro are identical. The available feature set is very comprehensive and conveniently categorized in exploration, synthesis, and evaluation procedures.

Numerical meta-analysis features
Lite
Pro
Exploration    
Data set overview
O
O
Data set summary (studies, N, event scarcity, etc)
O
O
Heterogeneity statistics (Q with P, I^2 with CI, tau^2 with CI)
O
O
Synthesis    
Fixed effect models
O
O
- Inverse variance
O
O
- Mantel-Haenszel
O
O
- Peto
O
O
Random effects models
O
O
- DerSimonian-Laird one-step
O
O
- DerSimonian-Laird two-step
O
O
- Paule-Mandel
O
O
Binary outcome data    
- Odds ratio (OR)
O
O
- Risk ratio (RR)
O
O
- Risk difference (RD)
O
O
Continuous outcome data measures    
- Mean difference (MD)
O
O
- Hedges' G
O
O
- Cohen's D
O
O
Generic data entry measures    
- Any measure with its standard error
O
O
Confidence intervals for synthesis result    
- Z-distribution
O
O
- T-distribution
O
O
- Bootstrap
C
C
Continuity corrections for non-event groups    
- Custom constant value
O
O
- Arm-dependent values
O
O
Cumulative synthesis
O
O
Bayesian synthesis
O
O
Evaluation    
Exclusion sensitivity assessments
O
O
- Single study exclusions
O
O
- Subgroup analyses
O
O
Dissemination selectivity statistics
O
O
- Rank correlation tests (Begg & Mazumdar, Rucker)
O/C
O/C
- Regression tests (Egger, Macaskill, Peters, Harbord)
O
O
Dissemination bias analysis
O
O
- Regression methods (Peters, Egger)
O/C
O/C
- Trim and fill analysis
C
C

O = available, X = unavailable, C = under construction

Basic graphical meta-analysis (Lite & Pro)

The basic graphical meta-analysis possibilities in MIX 2.0 Lite and Pro are identical. MIX 2.0 has the largest arsenal of graphs of all meta-analysis software. The available graphical options are not only very comprehensive, the graphs can also be customized easily via the right-click dialog boxes of Excel 2007. Alternatively, the graphs can be exported to PowerPoint 2007 and, with the vector-based graph items ungrouped, the graph can be adjusted in any way imaginable. The vector-based graphs from MIX result in high-quality imagines no matter how large they are made (the vector-based properties prevent pixelation) and are therefore ideal to create publication-quality graphs. Most journals accept such PowerPoint vector-based image files directly, but they are also easily imported and adjusted in graphical software such as Adobe Photoshop, Adobe Illustrator, or free alternatives such as Gimp and Paint.NET.

Graphical meta-analysis features
Lite
Pro
Exploration    
Simple forest plot
O
O
Histogram
O
O
Normal-quantile plot
O
O
Galbraith plot
O
O
L'Abbe plot
O
O
Heterogeneity funnel plot
O
O
Baujat plot
O
O
Synthesis    
Synthesis forest plot
O
O
Cumulative forest plot
O
O
Bayesian triplot
O
O
Evaluation    
Exclusion sensitivity plot
O
O
Weighting sensitivity plot
O
O
Selectivity funnel plot
O
O
Selectivity box plot
O
O
Selectivity regression plot
O
O
Trim and fill plot
C
C

O = available, X = unavailable, C = under construction

Meta-regression (Pro - under construction)

MIX 2.0 Pro can be used to model the outcome of studies as a function of up to 2 independent variables. However, the meta-regression facilities are based on fixed effect least squares modeling and must therefore be seen as explorative and hypothesis-forming. For more comprehensive meta-regression techniques, MIX 2.0 Pro provides a graphical user interface to logistic and random effects modeling in R and OpenBUGS.

Sample size calculations (Pro - under construction)

Sample size calculations for primary studies is often based on the assumption that the goal is to have sufficient power to detect a signal in the specific study itself. However, it is also possible to assume that the goal is to detect a signal via a meta-analysis updated with the planned study. This takes existing evidence into account. For health care professionals involved in meta-analysis this is a natural approach as the goal of additional studies is to provide a meaningful contribution to the existing pool of evidence. The sample size module in MIX 2.0 provides meta-analysts with the option to assess the impact of future studies with various kinds of sample size and effect size characteristics.

R and OpenBUGS code creators (Pro - under construction)

Although MIX 2.0 Pro is extremely versatile and comprehensive, it remains an add-in for Excel. This means that certain statistical approaches can be done faster and with more rigor in packages specifically made for statistics. R and OpenBUGS are two of such packages. R is a very versatile general statistical package and OpenBUGS is a package specifically made for Bayesian analyses using Gibbs sampling. Conveniently, both are open source and free and have a very active developer community. Unfortunately, both programs have a rather steep learning curve and MIX 2.0 Pro will provide an interface to produce code for meta-analyses in R and OpenBUGS.