Choose Language Hide Translation Bar
ceecee
Level III

JMP Can Be Shiny Too (2021-US-EPO-929)

Level: Beginner

 

Caroll Co, Research Scientist, Social & Scientific Systems

Helen Cunny, Toxicologist, Division of the National Toxicology Program, NIEHS

Keith Shockley, Staff Scientist, Biostatistics & Computational Biology, NIEHS

David Umbach, Staff Scientist, Biostatistics & Computational Biology, NIEHS

 

The dashboard functionality in JMP is a fantastic tool that can serve as a good alternative to R Shiny. Compared to R Shiny, JMP dashboards are easier and faster to build and can save lots of time in app development. In this poster, I talk about how JMP dashboards can be a tool for:

  •          Sharing the excitement of data discovery with others.
  •          Personalizing the presentation of graphics.
  •          Creating a visual storyboard to convey information.

 

 

Auto-generated transcript...

 


Speaker

Transcript

Caroll Co Hi, everyone. I am Carol Co. I'm a statistician at Social & Scientific systems, and today I'm going to talk to you about
  how you could use the dashboard functionality in JMP to showcase your data and to create a visual storyboard.
  So, unlike R Shiny, JMP allows you to create an app in a matter of minutes, without having to learn to code.
  So you could spend a lot more time in exploring your data and to do your statistical modeling, rather than coding an application from scratch. And I think that's really like the biggest advantage JMP has over R Shiny.
  So there are many reasons why an app can be useful. In my case, I was dealing with a complex data set
  with too many what-if scenarios and a lot of higher order interactions, and so producing static graphs had become too cumbersome and inefficient to work with.
  I was also working with the work group coming from different backgrounds and different technical and computational skills, so I was really looking for a platform where any user from any...coming from any background can find it easy to navigate.
  So now I'm going to show you how to build a dashboard using JMP. I'll switch to, like, my JMP data table.
  Okay.
  So here is a data set...a sample data set that came from a simulation study.
  I have eight columns in here, six of which are factors from an experiment and then I have two different responses that were collected. So because I've already spent a lot of time
  analyzing this data, I already know, like, some of the data features and what I want to go on my dashboard. So they are
  (let me show a panel here) like the distribution of all of my columns.
  Because I have two responses, I also wanted to show the relationship between my two responses and, as you can see, one of the...one of the factor in my data, which deals with whether the data is balanced or unbalanced, completely
  explains the separation between these two points. These two are features. And then I also wanted to show a graph builder to kind of link the relationship between my response and the different factors in my experiment.
  So, to build a dashboard all you need to do is go to file, click on New,
  and then choose dashboard. And JMP already has some built-in templates in here that you could use. If none of this really fits what you're looking for, just pick any one of them, because it's really easy to change them.
  So I already have kind of like a vision in mind of how I want this dashboard, this application to look like. And say I want the distribution to kind of go up here in the top right,
  my scatter plot to go here on the bottom,
  and then, this graph builder chart to kind of go like just right next to it.
  I also wanted to have a data filter in there, because of that, like, data balance issue that I...that I'm seeing with my scatter plot.
  So I forgot to mention that before you do this, like, the open reports that you have
  in your JMP session, which show up here as thumbnails in these...in the Left panel here, and so really all you're doing is, like, you're dragging and dropping to arrange...
  to have the layout that you want. And once you're happy with that, just click this Run Script icon in here. Just wait a couple seconds and here's your dashboard.
  This is basically what I intend my user to see. I don't want them to see the data table. This is...I don't want them to bypass that, I just want them to see this application so that they can start exploring the data.
  So for...
  for you guys who are familiar with JMP, you know that the...
  a lot of, like, the the interactivity, the dynamic linking is preserved and so that's really a great way to kind of showcase your data and tell a story.
  In my particular case, because we were dealing with two different responses, maybe one of the questions we're looking at was what are the settings that are giving us the optimal response.
  And so, in my case, in this case, it was responses where both response 1 and response 2 are greater than 80%.
  So if I just highlight that region, I can kind of like quickly see that most of these observations are coming from the higher sample size and when Factor C is levels A and B and so on.
  On the bottom right,
  again, because I've only had...I'm only highlighting the observations that are giving me the good response, I can kind of quickly see these three cells on the bottom left, none of them are highlighted so suggesting that the the settings that are in here are all poor choices.
  Another way to kind of like look at your data is to use, like, this filtering option. And because of this data feature that I have, I'm going to filter it by the data balance.
  And so now, if I just want to look at the end...the unbalanced case first, I can just quickly select that and everything is interactive and responsive.
  One more thing that you might be kind of like thinking, what is this stuff that's happening right here? This looks really weird. So
  you can kind of quickly highlight that region as well, and you can kind of see that the observations that are coming here are coming from the 100 sample size and when Factor D is Levels 3 and 4, and when Factors C is at E and D.
  So that's a really great way to kind of like explore your data.
  So as a data analyst,
  a lot of the,
  like, requests I get would be, like you know, changing the aesthetics of the graph.
  So, usually it would be somebody who doesn't really like my color scheme or like the line type that I chose, or even the arrangement of some of...because I have like, 1, 2, 3, 4...4 different factors in this plot.
  What's so great about presenting your data this way is that the user has complete autonomy in making those changes themselves, just by right clicking. So if you're familiar with Graph Builder, you know this, you can change the colors,
  the style and the width, but this...whenever I show this to other people, where you can, like, swap variables, so say if I want to swap this Factor E with...
  let's say, with sample size, you can get a completely different picture. And whenever I show this to people they're just like so stunned, because this is such a great tool to showcase your data.
  So if you're happy with the layout of this, I would suggest you just like hit this red triangle button up here.
  And then just save your script to your data table. Or if you're realizing there's something in here that you're not...you don't really like, you want to make changes, or you want to make it
  kind of more statistically savvy, maybe add a fit model...results from a fit model, you can go back to edit dashboard.
  And that'll take you back to your dashboard builder, where you can, like, take out some of these containers and add new ones or totally recreate it. Or you can have multiple versions of the same data set to cater to different types of audiences.
  So in conclusion, I think displaying your data in this way can be a really powerful tool to communicate your findings to your audience and
  your users, because of, like, the ease of use. Your use...your users
  can play with the application without even needing to touch the data table or needing to code anything.