The environment

The MIX program works in the environment of Microsoft Excel 2000 or later, currently only on PCs running Windows. It comes with a standard installer / uninstaller (.exe) and a simple click will bring the files to a (program files) directory of your choice. After making sure Excel has recognized it as add-in, you can run the program via an icon on your desktop or from within Excel itself.

The interface

Like a normal Excel session, the MIX program uses a menubar and several toolbars, which replace the standard bars that you will use normally in Excel. Your standard settings are remembered, and are restored once you exit the program. Several menus have menu items or submenus with submenu items. Most of the options in the menubar can be accessed directly via the buttons or dropdown menus in the toolbar, like for instance choosing an outcome measure, weighting method, or a modeling assumption. The program uses standard Excel sheets on which the output (numerical or graphical) is displayed. Since the whole interface with its menus are custom made, the MIX program feels like an Excel-type stand-alone program, although it really is just an add-in exploiting Excel's environment.

The data sets

The MIX program has 21 built-in data sets. At the moment, the data sets correspond to those being used in three popular books about meta-analysis in health care ("Systematic Reviews in Health Care: Meta-Analysis in Context" by Egger et al, "Meta-analysis in Medical Research" by Sutton et al, and "Systematic Reviews in Health Care: A Practical Guide" by Glasziou et al). The availability of these data sets in the program allows users to redo the analyses while reading the books experience for themselves what influence certain decisions have on some of the results. In the program, the data sets are conveniently sorted per book and can be loaded and unloaded with a few simple clicks. We intend to expand the data sets according to the wishes of the users, so if you have any recommendations, please let us know. Of course, you can also create your own data sets via manual input or import via CSV files and learn or teach with your own data.

The calculations

The numerical output is quite extensive for an Excel-based add-in. Currently, the output from the MIX program is, with the right settings, identical to the results from commercial programs. Below is a short list of current calculational options:

  • Input of descriptive & comparative data
  • Comparative data weighted by inverse variance method
  • Data sets with dichotomous & continuous outcomes
  • Meta-analysis outcome measures:
    • Risk difference
    • Risk ratio
    • Odds ratio
    • Mean difference
    • Hedges's g
    • Cohen's d
    • Correlation coefficient
    • Fisher's Z
  • Standard and cumulative meta-analysis
  • All meta-analysis results with CI and z-score with p-value
  • Fixed effect, random effects, and quality effects modeling
  • Cochran's Q with p-value
  • Higgins's I2 and H with CI
  • Several publication bias tests:
    • Rank correlation (tau-b) test with z-score and p-value
    • Egger's regression test with CI
    • Macaskill's regression test with CI
    • Trim and fill sensitivity analysis

Meta-regression, likelihood-based methods and Bayesian approaches are not integrated in the program and are currently not on our priority list.

The graphs

The MIX program can produce lots of graphs with just a simple click of a button. Many of the included graphs are not integrated in commercial meta-analysis software. Some are essential for certain types of analyses and others are merely helpful in data exploration or learning about certain concepts and calculations. Below a short summary:

  • box plot
  • weighted box plot
  • standardized residual histogram
  • normal-quantile plot
  • standard and cumulative forest plots
  • annotated forest plot (RevMan style)
  • p-value function plot
  • four types of funnel plots
  • funnel regression plot (Egger / Macaskill method)
  • exclusion sensitivity plot
  • Galbraith plot
  • L'Abbe plot
  • Baujat plot
  • trim-and-fill plot

Axis types and dimensions can be set for most graphs and in dot plots the studies can be labeled with and sorted by the study's ID, weight, reference number, sample size, p-value, control group value, effect size, date and quality score.

Analytical preferences

The MIX program can analyse data sets based on various user preferences. You can also apply advanced methods of continuity corrections, switch event codings, in- and exclude studies for subgroup analyses, use modified weights, and adapt various settings for rank correlation tests, regression tests, and variance calculations and assumptions.