Forest plot add in bug?

[#1]

Hi everyone,

I'm attempting to create a forest plot figure based on multiple logistic regression models.

I have 9 independent variables and four IQ outcomes, VCI, POI, PSI and Full IQ. Each outcome has three thresholds <70 vs >=90, 70-79 vs >=90 and 80-89 vs >=90.

All was fine but I detected some bugs when drawing the lines in the plot. For example, in females for Perceptual organization index (POI), results don't match with the ones at the database. At threshold 80-89 have a Lower bound lower than 1 (0,932086) and it always draws it as being higher than 1. There are other results like this one.

 I was puzzled with this and found some other ones that also do not match. However, most of the other ones on other variables match.

 I have been struggling with this for the past week, without any success.

I attach the database, a doc file that explains how I built the figure and a pdf with the final Figure, with the mistakes.

I found this to be very strange, because I checked the values in the dB and they are all correct.

Please help. I'm on a countdown for my PhD.

Best regards

Edgar Mesquita

[#2]

Hi,

I think that I have recreated the problem that you see.

One thing that I could not find in your description is that you added the variable "IQ" in the Group X role. I am guessing that you reverted to the Graph Builder control panel to do this as it does not appear to be a capability in the addin.

That seems to be where the addin has problems. I don't think that it was designed to support adding a Group X variable.

I took a subset just for IQ = 2 Perceptual organization index and ran the forest plot and it now looks as it should.

I think you will need to create separate forest plots for each category of IQ. Or you might be able to build the full plot that you want from Graph Builder.

I have attached the subset data table.

I hope this helps and good luck with your PhD submission.

Phil

[#3]

Hi Phil,

First, I want to thank you for providing the forest plot add-in. It is a great help for my recent project. 

I also ran into a problem with the add-in when I subset a dataset and only plotted a subset of the data. As you can see from the legend, in the wrong plot, it plotted the lower confidence interval, instead of the point estimate (I don't know how the CI was plotted when 95%LCL is plotted as the center point). If I created a separate data from the main dataset, then I got the correct plot. 

It is such a nice add-in. I hope the bug can be fixed. 

Thanks,

Jianfang