Suggestions to improve graphs

Everything related to development of the jamovi platform

by RCL » Thu Apr 25, 2019 4:59 pm

Hi. Congrats on jamovi again. It's become a wonderful stats package! :)

In the past, I was mostly using RStudio, SPSS and JASP in my daily work as a neuroscientist. I'm an intermediate user of stats, so I guess I'm the type of user jamovi's development has mainly focused on. Little by little, I've forgotten about other stats packages because jamovi has gotten so high-quality, comprehensive and efficient. However, I think jamovi needs further improvement when it comes to creating graphs. As an intermediate stats user who'd love to do everything in jamovi, I think several of these suggestions are essential for all intermediate users. What's the point of moving from another statistical software to jamovi if in the end you can't work with the graphs that jamovi creates? I think these suggestions should address the matter:

Essential improvements:

1. More plot themes: More are needed, specially themes that take into account common requests by journal editors. For instance, an alternative to the Hadley theme with grid lines instead of a solid grey background would be fantastic (I've attached an example). Many editors ask authors to avoid coloured backgrounds in order to save ink.

example.png (25.9 KiB) Viewed 2302 times

2. Size consistency: Something I find quite annoying is that plots' sizes and proportions are determined by the number of factors and levels the statistical test had. For instance, a RM ANOVA with a factor A (4 levels) and a factor B (7 levels) generates a figure that has a different proportion than a RM ANOVA with a factor A (2 levels), factor B (2 levels) and a factor C (7 levels). I understand it makes sense from a graphic point of view, but it would be great if it were possible to have some settings to play with. Sometimes you need to create a figure where one graph is a 2 x 7 RM ANOVA and the other graph is a 4 x 7 RM ANOVA. In jamovi, graphs have different proportions which means they won't look good on a paper. Then, either you have to take the plot into Photoshop or plot the figures elsewhere.

Highly recommended improvements:

1. Line patterns and colours: Every time I create a graph in jamovi, I end up taking it to Photoshop in order to change line patterns and colours. For example, if you are studying how fast people can visually detect emotional faces shown in different orientations (upright or inverted), you'll probably want to use a colour for a specific emotional valence and associate a line pattern to a particular orientation. You can't do that with jamovi. Adding options so the user can easily edit line patterns and line colours would be really helpful.

2. Bar charts: It would be very useful to have the option to plot descriptive data using bar plots. Don't get me wrong. Line charts in jamovi work splendidly. But sometimes you need to plot data that are the result of a very simple test. Like a t-test or a one-way ANOVA and a line chart feels a bit too complex for such a simple analysis.

Non-essential but practical improvements:

1. Within-subject error bars: This is really non-essential as it's a very easy to do. Sometimes when you get a significant effect error bars don't really help visually as they include between-subject variability as well. From time to time, it's more practical to plot within-subject error bars as described here: ... error-bars

Thank you so much for such an amazing statistical package. I hope my suggestions will be useful! ;)

Posts: 13
Joined: Sat Apr 21, 2018 11:56 pm
Location: Edinburgh, United Kingdom

by jonathon » Sat Apr 27, 2019 12:53 pm

hi renzo,

thanks for taking the time to put together these suggestions. we'll definitely have a think about the issues you've raised.

with thanks
User avatar
Posts: 1377
Joined: Fri Jan 27, 2017 10:04 am

by RCL » Mon Apr 29, 2019 3:36 am

Thanks Jonathon!
Posts: 13
Joined: Sat Apr 21, 2018 11:56 pm
Location: Edinburgh, United Kingdom

Return to jamovi development