scatterplot improvement

Discuss the jamovi platform, possible improvements, etc.

by adil.saribay » Sat Mar 06, 2021 10:07 pm

Hello all,

I've started teaching undergrad stats with jamovi and it's going pretty well and I'm generally very happy that my department has made the switch. I hope this is the right place to bring up a potential improvement.

I am teaching correlation this week and as far as I can see there is no easy way to produce a scatterplot matrix apart from the Correlation Matrix menu under Regression. But one can't remove the regression line and the error bars. The reason I would like to turn these off is because I don't want the students to get distracted, before they learn what these are. And in actual research, it's possible that one wants to display a scatterplot matrix without these elements.

There is also no way to get multiple scatterplots outside of this menu. I mean, if I wanted to produce multiple scatterplots at once (and not necessarily as a matrix, but just as separate images), I don't see any way. In JASP for instance, there is an option called "scatter plots" under descriptives that does this. And it's possible to turn off the regression line.

There is of course the scatr module but there, too, you have to drag the variables into the x- and y-axes and produce one scatterplot at a time. You can't select 5 variables and produce scatterplots for all possible pairs.

So, maybe this can be an improvement for scatterplots; i.e., being able to produce multiple separate scatterplots by selecting more than 2 variables at once and having the option to not have the regression line and error bars in such plots.

cheers!

P.S. the jjstatsplot module has just come to my attention but I think it has the same problem: No way to get multiple scatterplots from one menu. By the way, I'm not sure if we should refrain from making a feature/improvement request for core jamovi if there is already a module that has the requested feature/improvement. Is core jamovi trying to incorporate features found in modules or is the strategy more to let the community extend jamovi with these modules?
adil.saribay
 
Posts: 2
Joined: Sat Mar 06, 2021 9:48 pm

by jonathon » Sun Mar 07, 2021 3:03 am

hi,

Is core jamovi trying to incorporate features found in modules or is the strategy more to let the community extend jamovi with these modules?


a bit of both. we like the community to extend with modules, and good ideas we like to fold back into jamovi core.

i think we'd like to improve the correlation matrix, at the moment we're using the ggally package, and we'd like to write something from scratch, but it's not a small undertaking, and the correlation matrix is in many ways "good enough".

the challenge is always striking a balance between providing lots of features, but not over-cluttering things. it's a delicate balancing act.

jonathon
User avatar
jonathon
 
Posts: 1689
Joined: Fri Jan 27, 2017 10:04 am

by jonathon » Sun Mar 07, 2021 5:31 am

but i think this illustrates a maxim in statistical software ... it doesn't matter what you provide, someone always wants something else :P

thanks for bringing this up.

jonathon
User avatar
jonathon
 
Posts: 1689
Joined: Fri Jan 27, 2017 10:04 am

by adil.saribay » Sun Mar 07, 2021 4:59 pm

It's only a suggestion of course and it may not make sense or the community may not like it overall. I'll love jamovi regardless :-)

I completely understand the importance of keeping the balance between feature-richness and ease-of-use. When I think of that, it makes sense that the core remains minimalist and uncluttered while the modules add whatever people want. A minimalist core makes it easier to teach stats at the most basic level. For that reason, maybe I should not make such improvement suggestions unless I'm super compelled.

Thanks for the reply Jon!
adil.saribay
 
Posts: 2
Joined: Sat Mar 06, 2021 9:48 pm


Return to General