Using graph builder to identify data, then making subset of the identified data: unique problem

[#1]

Good Day,

I am sure there is a way to script my way out of this issue, but I am attempting to do it without writing script. JMP community input appreciated!

The data are from a particle movement model that I put through a hydrodynamic model (river). I like to plot the path of the particles using coordinates to determine where the particles go in 2D. The paths are broke into 20 second increments of model time so it is essentially a running times series of particles responding to model parameters as they progress through the modeled river system Doing it this way allows me to calculate variables of interest for each particle (time, velocity, etc). I release 500 particles per scenario and let them run through the model (river). The output when plotted with graph builder shows the particles path through the river.

Problem statement: I want to identify the particles that went through off-channel sections of the model space separately so I can compare the entire run length/path of particles that went through off-channel sections to those that stayed out of the off-channels and in the main channel.

I typically use the "Lasso" function to grab the subset of particles I am interested in for analysis, but this time I want find the particles that used the off channel areas and compare them to those that didn't across the entire run of the model. The lasso function constrains the data to just the section highlighted, not the entire run of the model.

Below you will see the lasso function being used to highlight the areas in pink I am calling "off channel". The black in the below are the particles path plotted with their X,Y coordinates every 20 seconds of the model run time.  I want to identify and label the entire run of each particle across the model run, not just the section highlighted so I can separate them from the rest of the particles that stayed in the main section of the river. 

Attached are a portion of the data. The real data file is too big to attach.  The column "Scenario"  denotes four different conditions of the model, plotted above and "wrapped" by scenario to show that particles respond differently under each scenario. In the attached I have selected just 2 scenarios (07_225_CH0 & 07_225_CH1) to make the file smaller. The particle unique identifier is the column labeled "FID" which corresponds to the 500 particles I release per scenario. So each scenario has an FID == 1-500. I want to use graphics to identify which FID particles went through the "off channel" areas and which stayed in the main river so I can analyze the effect of off-channel migration routes to particle parameters. I plan to do this many times in the coming year and hope to find an easier way of doing it besides hand entering a unique identifier for each FID by scenario. 

Any thoughts or input appreciated.

JMP V 16.1

Thank you!

[#2 · Accepted solution]

Is this what you're looking for? It's faster to do than to explain

1. Make sure nothing is selected in data table
2. Select a portion of one of the off-channels
   1. 
3. In the data table, select the Scenario and FID columns
   1. 
4. Right-click in the row header space on the left of the data table, select "Select Matching Cells" 
   1. 
5. Now the entire off-channel subset is selected
   1. 

[#3]

Thanks for engaging Jed!

Yes, that seemingly selects all of them.... Doh.... I spent most the night trying to figure that out....

This support page is awesome!!!!!!!!! Look forward to supporting it once I learn the system better. 

Thank you so much!

[#4]

Hi @Frogger,

I might suggest an alternative to using Graph Builder to begin visualizing the data.  Maybe use a bubble plot.  

Graph>Bubble Plot

Maybe setting similar to the following--

Bubble Plot launch dialogue settings

This way you can follow particular FIDs across time for the off-channel behavior.
From the red-triangle I turned on Trail Lines for selected.

Red Triangle>Trail Lines>Selected

Bubble Plot animation

And then perhaps use Graph Builder in a second step as noted by @Jed_Campbell 

cheers,

Stan

[#5]

Absolutely. Good call @stan_koprowski 

That will be helpful for visualizing individual paths that are problematic. I will explore now and see if I can figure it out. 

Every time I think about how to look at these data JMP has a solution. Amazing product. 

[#6]

@Frogger   If it is reasonable to make all of the X coordinates consistent by rounding or binning, then this enables converting each trajectory into a time series with the new X serving as a surrogate for time.   This opens up several new and powerful ways to analyze the data, for example:

- Transpose the data to have one column per X value, then use Hierarchical Clustering to construct a heat map and dendrogram of the trajectories.  The off-channel ones should be readily identifiable, clustered, and selectable by clicking the associated tree branch

- Functional Data Explorer

- Time series analysis with Time Series, Time Series Forecast, XGBoost, or the new Torch Deep Learning add-in (requires JMP 18 Early Adopter)

[#7]

@russ_wolfinger,

Thank you for following up with this idea.  An interesting approach, but I fear by clustering the x coordinates I will loose specificity of their position and not be able to differentiate the off channel from main channel fish. Especially when they are close to each other.  Maybe I am not understanding your approach?