We've tried to make Meta-Analyst as intuitive as possible, and hope that most operations are straight-forward even without a manual. We also provide contextual help within the program explaining most of the basic functionality and options. This document is meant to supplement the contextual help, and concentrates on some of the newer tasks that we see as being non-obvious.
Table of Contents
Working With Data
Meta-Analyst uses a custom XML-based data format, .ma, which is a simple combination of csv data for the rows comprising a dataset and some meta-data (e.g., covariate and label names, data type, etc.). You can enter data by hand, paste it directly from Excel, or import from files formatted as Excel (.XLS) or Comma Separated Value (.CSV) files. Once you've imported your data once, it will be saved to a .ma file which you may subsequently work with, using File →Open Dataset.... This section covers how to manually enter or import data and how to work with covariates and labels. As a general rule, data manipulation can be done via right-clicking. For example, to delete an existing row from data, you would right-click anywhere on that row and select "Delete Study" from the context-specific menu that appears.
Entering Data Manually
To enter a dataset by hand, you'll want to select File → New Dataset... and then either Binary, Continuous or Diagnostic, contingent on what sort of data you'll be entering. If you'll be working with binary or continuous data, you'll also need to specify whether or not you're working with two arm (i.e., treatment and control) or one group (just treatment).
Once the type of data is selected, you will be presented spreadsheet in which data can be entered. The user can also add covariates and labels, and otherwise manipulate the data, via a context menu that appears when any column header is right-clicked (see figure below). Likewise, right- clicking anywhere in the spreadsheet will pop-up a different contextual menu that allows you to delete a study or copy data.
If you want to delete or exclude a covariate or label, right-click on the corresponding column header and select Delete Covariate (or Delete Label). You can also exclude covariates and labels in this fashion.
Importing Data
Meta-Analyst can currently read Microsoft Excel (.XLS) and Comma Separated Value (.CSV) data. Note: in order to import Excel files, you will need to have Excel installed on your computer. At least one user has reported being unable to import Excel data directly, even though Excel was installed on their computer. This is being investigated, but in the mean time, you can always save your Excel data as a .csv and then import it (or, you can copy and paste it directly, explained in detail in a bit).
Let's walk through a sample import. Suppose we have the following data in Excel, saved in the file "excel_data.xls":
Further suppose that columns C and D are the treated events and treated sample size for the control group, respectively. From within Meta-Analyst, you then click on File →Import Dataset.... You will see an Excel-like spreadsheet with columns. Here you can drag and drop the columns (just click on a column header, hold, then release in the position you would like the place the column) to fit your data. For example, the 'default' in Meta-Analyst for binary two-arm studies has the treated group fields followed by the control; in the figure below this ordering has been transposed.
Meta-Analyst will now import the data in the correct order. The import process is the same for .csv files; just browse for files of type ".csv" (CSV Files) when looking for the file containing your data.
If you'd rather, you can copy & paste your data directly from Excel. For example, in the figure below I have moved the columns around in Excel so that the control group figures are first and then copied them (using ctrl+c or by right-clicking).
Next, Simply right-click on the column you want to start pasting into (in this case, 'Treated Events') and select "Paste Excel data" from the pop up menu (alternatively, just use ctrl+v).
Labels and Covariates
The main distinction between labels and covariates is that the latter will be used in the analysis, when applicable, while the former is never used in computation (it's just a label!). At the moment, Meta-Analyst can only handle numerical covariates. We plan on providing automatic mapping of nominal covariates in the future, but for the time being you will have to either exclude nominal covariates, encode them numerically with a suitable mapping, or else use them as labels.
Labels and covariates can both be added by right-clicking any column header in the data spreadsheet and clicking "Add Covariate...". You will be prompted for a name, and once entered, a new column will be added to your data (either immediately before or after outcomes, for covariates and labels, respectively). Once added, you can edit existing covariates and labels (e.g., change their names, exclude and delete them) using the contextual menu available via right-clicking on the corresponding column header.
At the moment, labels are parimarily useful for two things: subgroup analyses, in which case they can be used to group your studies; and forest plots, as label columns will be displayed on forest plots (see figure below, where two labels were added to our data).
Performing Analyses
In this section we outline how analyses (including basic and advanced meta-analysis, and explatory plots) are performed in Meta-Analyst. We also demonstrate how forest plots can be edited.
Meta-Analysis
Standard meta-analyses can be performed by selecting Analysis → New Analysis... from the drop down menu. You will be prompted with a dialog as shown below. The analysis will be run over all data in the current spreedsheet that has not been excluded via the corresponding checkbox. From the dialog box, you can specify: The output directory to which analysis data will be saved, the name of the analysis to be run, and, most importantly, the model for the analysis and its corresponding parameters. For example, if you are running a fixed effects meta-analysis, you will here select which fixed method (e.g., Peto) is to be used.
Once you've specified the details, clicking "OK" will start the analysis. Note that if you have selected a "simple" meta-analysis -- for example, random or fixed effects -- analyses will be run for every available outcome metric (e.g., Odds Ratio, Risk Ratio and Risk Difference, in the case of binary data). The figure below shows how the results will be rendered once the analyses are complete. Clicking on a node in the right side tree navigation will scroll the corresponding table or graphic into view.
Cumulative Meta-Analysis
Cumulative meta-analysis can be performed over all but diagnostic data (currently). To perform a cumulative meta-analysis, select Analysis →Subgroup Analysis.... You will be prompted with the standard analysis details dialog (the analysis specified here will be applied at each step in the cumulative analysis). Once your specifications are entered, press OK, and results like those shown below will be displayed.
Subgroup Analysis
Subgroup analysis can be performed over any type of data. To perform a subgroup analysis, you need to have at least one label column in your data. You can then click Analysis →Subgroup Analysis.... This will bring up a dialogue asking you to specify which label to group the studies by, and what field to sort by within these groups. Once these are selected, you will be prompted with the standard analysis dialogue asking you to specify the type of analysis to run (and select the corresponding parameters). This analysis will be run over each of the subgroups separately and ultimately over all of the data combined. The figure below shows an example sub-group analysis forest plot using the same data in the above forest plot figure; we grouped studies by "Label 1" and sorted within these groups by "Label 2". You can also edit the subgroup plot by right clicking the plot and select "Edit forest plot...". We will next explain more about editing forest plots.
Editing Forest Plots
Once an analysis is run, you can edit forest plots by right-clicking on the graph and selecting "Edit Forest Plot". The first thing you see in the forest plot editor is the column ordering. Here you can drag-and-drop columns to change their display order. For example, in the figure below, we have dropped "Label 1" and "Label 2" on the left side of the plot. You can also right-click on the column headers (within the editor) and choose to hide them.
The other tabs on the forest plot editor allow you to set the scale, change the symbols, etc. Play around a bit and see what you can do.
Right now, once the plot is edited, you can copy and paste it into other programs (e.g., Word), or save it as a stand-alone image (both of these can be done via the right-click menu that appears when you right-click the forest plot). However the output (PDF and Word files) generated automatically during the analysis will not be updated. There is not currently a way to re-open edited forest plots, nor is there a way to save defaults for your forest plots. We are working on these features.
Meta-Regression
Meta-Analyst can perform both fixed and random effects meta-regression. So long as you have a covariate in your data, the two meta-regression options will be available via the drop-down Analysis menu under the Meta-Regression sub-menu. Once run successfully, you will be presented with results similiar to those shown below.
Exploratory Plots
Meta-Analyst can perform treatment vs control effect and treatment effect vs covariate plots for continuous and binary outcomes (the former is available only for two group data; the latter only when the data has a covariate). Both of these can be found under the Analysis → Exploratory Plots sub-menu. Below is an example of a simple tx effect vs covariate plot generated by Meta-Analyst.
