blandr module development

Everything related to the development of modules in jamovi

by jonathon » Wed Feb 21, 2018 1:26 am

procrastination is how most jamovi modules get written :P
User avatar
jonathon
 
Posts: 869
Joined: Fri Jan 27, 2017 10:04 am

by deepankardatta » Wed Apr 11, 2018 5:18 pm

Hi all

I've updated the Jamovi module to something better. It can be found at https://github.com/deepankardatta/blandr/blob/master/blandr.jmo

I realise I've totally failed at getting GitHub desktop to not upload changed .JMO files, but actually it's worked out well - the forum doesn't allow uploading these files anyway!

The changes to it from the last time I've posted:
* updating of the underlying R package - this has changed some functions so that I can submit a 0.5 version to CRAN soon
* got rid of the options to change the plot title - I just couldn't get this to work without errors
* added options to plot for proportional bias

MAgojam (memberlist.php?mode=viewprofile&u=177) has contributed to the development and suggested some improvements for the future which I'll hopefully get round to soon: https://github.com/deepankardatta/blandr/issues/3

I was hoping that forums members could kindly test the module? Hopefully on Windows, Mac and Linux. If it works well then I'd be trying to see if Jonathon could add it to the Jamovi library

Thanks!
deepankardatta
 
Posts: 21
Joined: Mon Jan 29, 2018 4:30 am

by dropmann » Wed Apr 11, 2018 11:31 pm

Hi! good work with the module. It doesn't have any issues on windows and Ubuntu. We'll get back to you about macOS.
dropmann
 
Posts: 57
Joined: Thu Feb 02, 2017 10:26 am

by jonathon » Thu Apr 12, 2018 12:07 am

oh, this sort of module *just works* everywhere, so generally doesn't need testing on different OSes (but it does work on my mac).

here's some feedback (let me know if you'd prefer this on your github in future?)

so the bland-altman statistics outputs a lot of raw output, and it seems to output all the raw data too. this will become a bit unwieldy if the user has a large dataset. also, the bland-altman analysis is a bit sparse. how about combining these two analyses into one, and providing the statistics in nicely formatted tables?

wrt ui, try and avoid using verbs in option labels ('plot', 'clarify'). i.e. just 'proportional bias line'. if you feel like you want to make it clearer that this option pertains to the plot, you could put all these options under a 'Plots' heading.

but otherwise, looking good!

cheers

jonathon
User avatar
jonathon
 
Posts: 869
Joined: Fri Jan 27, 2017 10:04 am

by deepankardatta » Fri Apr 13, 2018 12:34 pm

Dear all

Thanks for the feedback. I've uploaded a new version for testing: https://github.com/deepankardatta/blandr/blob/master/blandr.jmo

(1) I still have failed to get Github Desktop to stop uploading my .jmo files, but a benefit for sending out test modules!
(2) Feedback here is fine!
(3) I've updated the table on the analysis to make it more readable and have more information - is there anyway to separate the columns visually?
(4) As 4 the analysis function should be the main function over time. I kept the statistics function (now titled raw statistics) so that I can do any error checking whilst making the module directly in Jamovi.
(5) Taken Jonathon's suggestion for UI clarity
(6) Is there an easy way of adding the dividers between Jamovi functions from the dropdown menu? From seeing the jmv examples you can do it on the u.yaml file, but then this just seems to get overwritten by the compiler.
(7) Suggestion - is it worth combining the a / r / u yaml files? I can see the point in separation of these concepts, but it makes developing a new analysis very unwieldy for people whose main job isn't programming.
(8) In the analysis menu, there is an option for marginal histograms. If you get it work on Windows do let me know! The underlying function works in RStudio, but doesn't work in Jamovi with marginal histograms enabled. This is also the case when you run the jamovi function directly in R. Not sure why, but could be that Mac OS Jamovi is on an older version of R. Not sure yet.

With Thanks
deepankardatta
 
Posts: 21
Joined: Mon Jan 29, 2018 4:30 am

by deepankardatta » Fri Apr 13, 2018 6:07 pm

Note points above 5 and 6. Somewhat fixed it by having a look through the JMV github for inspiration.

The module now has a divider in the drop down menu between the analyses. Also the analysis screen now has the checkboxes under a plotting options collapse box.

I still need to work out a better way to format the table, but I think it is nearly in good enough / "eat your own dog food" state.

Just wish I had something like Jamovi earlier on in my career!
deepankardatta
 
Posts: 21
Joined: Mon Jan 29, 2018 4:30 am

by Ravi » Fri Apr 13, 2018 6:38 pm

This looks nice! Just a quick comment (I will do some more elaborate comments at a later point): The main Bland-Altman table is very wide at the moment. How about something like this:
Screenshot from 2018-04-13 11-34-50.png
Screenshot from 2018-04-13 11-34-50.png (34.36 KiB) Viewed 2346 times
User avatar
Ravi
 
Posts: 140
Joined: Sat Jan 28, 2017 11:18 am

by jonathon » Fri Apr 13, 2018 11:20 pm

hey!

(1) add *.jmo to a .gitignore file, while you're at it, you can build/js and build/R and temp/

(3 and ravi's suggestion), take a look at table folding. i don't think i've documented this properly yet, but you can see it at work with the walrus module.

https://github.com/jamovi/walrus/blob/m ... r.yaml#L18

when you use the square brackets in column names, rows get folded down underneath

(6) you mean in the blandr module menu in the ribbon? it looks like you figured this out in your latest version, but let me know if i've misunderstood. (ribbon menu stuff is all handled in the .a.yaml)

(7) in lots of cases, these files become very large, and keeping them separate helps keep them manageable.

(8) what OS are you developing on? at present we only ship R 3.3 on macOS, because we're wanting to maintain compatibility with mavericks, but at some point we'll update it.

cheers

jonathon
User avatar
jonathon
 
Posts: 869
Joined: Fri Jan 27, 2017 10:04 am

by deepankardatta » Mon Apr 16, 2018 6:19 am

I'll try over the next week to update the table into a better format - in fact that's what I was thinking of, however the coding to do it was a bit beyond me. I'll have a look at the walrus module and see if I can work it in.

I'm on a macOS High Sierra setup.
deepankardatta
 
Posts: 21
Joined: Mon Jan 29, 2018 4:30 am

by MAgojam » Mon Apr 16, 2018 12:34 pm

deepankardatta wrote:Dear all

Thanks for the feedback. I've uploaded a new version for testing: https://github.com/deepankardatta/blandr/blob/master/blandr.jmo


Hi @deepankardatta,
if I install your blandr.jmo of 12.9MB in Win 7 pro or Win 10 Pro (ITA), I do not get output.
If you recompile it from RStudio (27.0MB), it works correctly (see screenshot).

blandr.png
blandr.png (73.44 KiB) Viewed 2324 times


@Ravi's suggestion for the main Bland-Altman table is a great suggestion to make it more readable.
Thank you for your excellent work.

Greetings,
Maurizio
User avatar
MAgojam
 
Posts: 49
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

PreviousNext

Return to Module development

cron