The Forest Plot add-in displays several confidence or credible intervals vertically. The user supplies the lower and upper limits, the center value (which could represent the mean effect) and labels for each interval from a data table. The user may provide a second center value to present additional summary values (e.g., medians) in the figure. Color options allow you to highlight intervals that exclude the null value (default 0), and this color varies according to whether the interval is above or below the null value. A dotted vertical reference line is drawn at the null value. Intervals are ordered in the figure according to their order in the data table.
Version 3 allows the user the specify a variable to indicate the size of the Center Value. The area of the center value is proportional to to Marker Size. Users can specify the minimum size for center symbols, apply a scaling factor to control symbol size, and a transparency factor so that narrow intervals contained entirely within the Center Value symbol are visible. This does not affect the sizing of the Center Value 2 marker, which has been changed from a diamond to a pipe to improve visibility. See the JMP blog for some examples of the forest plot with the newly provided nicardipine data.
The figure below displays the dialog for an example using summary statistics for 20,000 MCMC samples from a Bayesian hierarchical model of 40 adverse events from Mehrotra & Heyse (2004).
The figure below displays the 95% equal-tailed credible intervals based on the lower 2.5% and upper 97.5% quantiles. Using the sample data, interested users can create a similar plot for the highest posterior density (HPD) intervals by changing the lower and upper limits.
Mehrotra DV & Heyse JF. (2004). Use of the False Discovery Rate for Evaluating Clinical Safety Data. Statistical Methods in Medical Research 13: 227-238.
Version 4 fixes a bug when reproducing the plots when saved to a data table and adds a distinctive symbols when labels are either "total" or "overall". See below figure produced from RoseMeta.jmp. Center bubbles are sized according to the total count of subjects in a particular subgroup. In lieu of a bubble sized by number of subjects, the Overall confidence interval (representing the entire trial population) is displayed with an asterisk. Data is from a clinical trial in patients with type 2 diabetes to assess excess cardiovascular risk of rosiglitazone (Home et al., 2009).
Home PD, Pocock SJ, Beck-Nielsen H, Curtis PS, Gomis R, Hanefeld M, Jones NP, Komajda M & JJV McMurray. (2009). Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): A multicentre, randomised, open-label trial. Lancet 373: 2125–2135.
Version 5 adds two new features. The first feature is an option to add Confidence Interval Text. This allows the user to include a text string along the right axis. This could include text for a test of interaction for the particular subgroup or, as shown here, the estimate and 95% confidence interval for the hazard ratio for rosiglitazone versus control. The second feature is a check box to Reverse Order of Axes. This saves the manual step of having to reverse order for the vertical axes. Without the presence of this second feature, text provided using the first feature would require both axes to be reversed separately. With reverse order, the Overall group is placed at the bottom of the figure. When the option is unchecked, Overall is placed at the top.
Version 6 corrects a bug that would cause an error when the Forest Plot script was saved to a data table and rerun.
Version 7 adds a subgroup role to the dialog to more easily create subgroup forest plots with references lines, and an option to color Center Value 2 according to the selected color preference or not (which defaults to black). The resulting plot based on the above dialog selection is presented below.
Version 8 fixes a bug in JMP 13 when Marker Size option is used.
Hi Winfried. Sorry for the delay in response. The subgroup figure that was generated above for the Home et al study was generated in the following way. The dialog was run as above in the graphic. Then I click on the red triangle option to get the Graph Builder control panel, then I drag level to nest it on the y-axis. I need to drag out the x-axis to make it a bit nicer looking.
For the lines between subgroups, right click on axis and choose Axis Settings. Then turn on Show Grid on the second tab (Label Row 2). This will add lines to divide the subgroups.
Hi Winfried. I have turned off the option to hide the toolbar. It should be visible for you now in V5. The issue you are reporting with the JMP journal is likely due to the Add Graphics Script(). I can try to explore why this happens, but there may not be a way around it. However, I did notice that if the data table was open and current, reopening the closed journal would execute the plot correctly. It would be an additional step, but it may be possible to add some text to the journal file to open the data table and make it current. For example,
dt = Open("C:\Users\rizink\Desktop\rosemeta.jmp", invisible);
Current Data Table(dt);
Regarding the last point, the add-in works by adding a Value Ordering property to the data table for the variable listed in Label. This is done to easily maintain the order of the data table (or the reverse) in the Forest Plot. Otherwise, the default behavior orders rows alphabetically which would likely not give what is wanted. Below is a data table that had a Forest Plot created. The star on Subgroup (where the hand is) will show there is a Value Ordering on the variable. Clicking it will open the properties for this column, and Value Ordering can be removed by clicking "Remove" the in the dialog. Data can be sorted, and the plot regenerated. In general, the order should be determined before running and saving the plot.
