About FN

FN

Well, this is a bit embarrassing. I though I was submitting an idea and not volunteering to lead the session! In any case, my goal is to have a quick reference for those that know Python/Pandas syntax and want to replicate something similar in JSL. https://cheatsheets.quantecon.org/index.html https://cheatsheets.quantecon.org/stats-cheatsheet.html Is there something like this already published for JSL?

FN

I am trying to execute a .bat file from JSL but it does not work as expected. What am I missing? https://www.jmp.com/support/help/en/15.1/#page/jmp/run-external-programs.shtml rp = RunProgram( Executable( "cmd.exe" ), Options( {"C:\temp\myfile.bat"} ), ReadFunction("text") // returns all text ); //"Microsoft Windows [Version xxx] //(c) 2018 Microsoft Corporation. All rights reserved. // // C:\Program Files\SAS\JMPPRO\15>"

FN · ‎09-18-2020

I need to script the creation of a new column with the row number only. As this is part of an add-in, this column may exist already or the table may have the same column name by chance. This column can be hidden to the user but will be later used for virtual linking tables. What would be the best approach? Maybe to add a property to make this column unique?

FN · ‎09-18-2020

PythonSend() saves temporary csv which are read with the default pd.read_csv('myfile.csv'), hence no parse_date. To deal with this, one option which is not impacted by the date format: ClearLog(); dt = Open( "$SAMPLE_DATA/Functional Data/Weekly Weather Data.jmp"); t = dt << Subset( Allrows, Columns(:DATE) ); t:DATE << Format(best); Python Init(); Python Send( t ); // send the opened data table represented by dt to Python Python Submit( " import pandas as pd print(t.dtypes) print(t.head()) t.iloc[:,0] = pd.to_datetime(t.iloc[:,0], unit='s', origin=pd.Timestamp('1904-01-01')) print(t.dtypes) print(t.head()) "); Python Term(); // Output // DATE int64 // dtype: object // DATE // 0 3534451200 // 1 3535228800 // 2 3537648000 // 3 3538252800 // 4 3538857600 // DATE datetime64[ns] // dtype: object // DATE //0 2016-01-01 //1 2016-01-10 //2 2016-02-07 //3 2016-02-14 //4 2016-02-21 If there is a better way, let me know.

FN · ‎09-16-2020

Indeed, without context it does not make much sense. PythonSend() works by saving a temporary csv and loading it into pandas. When dealing with dates, current JMP implementation does not have the date_parser active and the object is load as text instead of a number. Python Send() does not recognize datetime format (sent as string) As my date column is within other X's which are continuous, I need a way to either filter that column out or transform it to integers.

FN · ‎09-16-2020

I can also change the date to integer (sum of seconds since starting date). Would it be better to change the format of all the columns in a data table that has already numeric values considering that I can have 100.000 rows and hundreds of columns?

FN · ‎09-14-2020

When sending datatables to pandas dataframes (Python), numeric values (datetimes) are sent as strings. I would not expect this behavior, since now I have to deal with this manually (considering datetimes in Python and JMP are different). Any workaround? I imagine the easiest would be to transform the JMP format to integer, send this information, and transform the date considering the logic behind JMP dates (sum of seconds since specific date). ClearLog(); dt = Open( "$SAMPLE_DATA/Functional Data/Weekly Weather Data.jmp"); t = dt << Subset( Allrows, Columns(:DATE) ); Python Init(); Python Send( t ); Python Submit( " import pandas as pd print(t.dtypes) "); Python Term(); // Output // DATE object // dtype: object // Types: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.dtypes.html //float float64 //int int64 //datetime datetime64[ns] << Expected //string object //dtype: object

FN · ‎09-11-2020

I a have list containing columns the user has selected. These are already numeric, but I want to find those that are with time or date format as they need special treatment. How can I filter these using JSL?

FN · ‎09-10-2020

Does virtual join create the interactive linking when selecting rows or columns? In the examples I ran this did not happen.

FN · ‎09-10-2020

I have a few hidden data tables that I only show to the user via some plots. As they are hidden, they accumulate in memory (as many times as the user calls the function in the toolbox). Is there a way to embed the JSL action (closing the invisible table) when the user closes the figure created?

FN · ‎09-10-2020

Yes, it was a workaround in my script to deal with this bug. https://community.jmp.com/t5/Discussions/Python-Send-skips-NaN-values-when-single-column-is-sent/m-p/285293

FN · ‎09-10-2020

What's the easiest way to duplicate elements in a list myListA = {"y"}; //Result I need: myList = {"y", "y"}; // This should work with an arbitrary number of elements in the list myListB = {"y1", "y2", "y3"}; //Results myList = {"y1", "y2", "y3","y1", "y2", "y3"};

FN · ‎08-12-2020

Thanks. So, as it seems this is not possible for JMP 15: Here is the proposal for those that are interested: Plotting packed distribution with equal width and no jitter

FN · ‎08-12-2020

When using violin plots (density), there is an option to have equal widths per variable. However, we are unable to plot the dot distribution with the same shape as it does not respect the width of each variable. // Start of script; Names Default To Here( 1 ); Clear Log(); dt = Open( "$SAMPLE_DATA/Bands Data.JMP" ); Data Table( "Bands Data" ) << Stack( columns( :proof cut, :viscosity, :blade pressure ), Source Label Column( "Label" ), Stacked Data Column( "Data" ), Drop All Other Columns( 1 ) ); Graph Builder( Size( 534, 456 ), Show Control Panel( 0 ), Variables( X( :Data ), Y( :Label ) ), Elements( Points( X, Y, Legend( 5 ), Jitter( "Packed" ), Jitter Limit( 0 ) ), Contour( X, Y, Legend( 6 ), Violin Scaling( "Equal Width" ) ) ) ) It should be straight forward to achieve something like this but for a number of variables.

FN · ‎08-11-2020

When using violin plots (density), there is an option to have equal widths per variable. However, I am unable to plot the dot distribution with the same shape as it does not respect the width of each variable. // Start of script; Names Default To Here( 1 ); Clear Log(); dt = Open( "$SAMPLE_DATA/Bands Data.JMP" ); Data Table( "Bands Data" ) << Stack( columns( :proof cut, :viscosity, :blade pressure ), Source Label Column( "Label" ), Stacked Data Column( "Data" ), Drop All Other Columns( 1 ) ); Graph Builder( Size( 534, 456 ), Show Control Panel( 0 ), Variables( X( :Data ), Y( :Label ) ), Elements( Points( X, Y, Legend( 5 ), Jitter( "Packed" ), Jitter Limit( 0 ) ), Contour( X, Y, Legend( 6 ), Violin Scaling( "Equal Width" ) ) ) ) I would like to know how to achieve something like this but for a number of variables.

FN · ‎08-11-2020

I believe I found an inconsistent behavior when sending datatables to pandas dataframes (Python). I would expect that if columns have missing values, these are passed as NaN to pandas dataframes. Yet, this is only true if more than one column is sent, otherwise Pythons only receives non missing values. ClearLog(); dt = Open( "$SAMPLE_DATA/Bands Data.jmp"); Explore Missing Values( Y( :caliper, :viscosity ), Missing Value Report ); x1 = dt << Subset( Allrows, Columns(:caliper) ); x2 = dt << Subset( Allrows, Columns(:caliper, :viscosity) ); Python Init(); Python Send( x1 ); // send the opened data table represented by dt to Python Python Send( x2 ); // send the opened data table represented by dt to Python Python Submit( " print('x1:', len(x1)) print('x2:', len(x2)) " ); Python Term(); // Output: // x1: 512 // x2: 539

FN · ‎07-30-2020

I have one a table opened by the user. After some python code I generate another table that ends: having a subset of columns (keeping same column names) sharing its time index (few rows might have been removed). How can link these two tables via JSL? The functionality I am after is to have something equivalent as if I did a subset in JMP. // Start of script; Names Default To Here( 1 ); Clear Log(); dt = Open( "$SAMPLE_DATA/Bands Data.JMP" ); dt << Subset( Linked, Suppress formula evaluation( 0 ), Sampling Rate( 0.9 ), columns( :timestamp, :proof cut, :viscosity, :caliper, :ink temperature ) )

FN · ‎07-28-2020

Looking at your solution again, if one normalizes each column variable first then you can color them directly within graph builder. Notice that the example worked because the ranges were similar. The concatenated table can have two columns to keep all the information: - Value (original) - Normalized value (used for color) The final question to accept the solution: How tall can a JMP table be? How to link with original table? The alternative for the second point will be to: a) show the original value as information when you are hovering your mouse b) or create graphlet with a column filter that changes the plot (image above). As this will be a very long table it should be a solution that does not make JMP struggle.

FN · ‎07-28-2020

It does seem to do the trick, thanks. Two small questions about this solution: 1. Can you script code the row colors? 2. Is there a way to dynamically link the two tables so when one value in the original column is selected the row of the concatenated is also highlighted?

FN · ‎07-28-2020

No problem Franck. It is a great contribution, thanks for sharing.

FN · ‎07-26-2020

I would like to plot the data distributions but coloring each variable individually. Example: I found the following property which encodes color base on max/min gradient or user specific data. However, this might not the be the right approach. The code for the example which I manually composed to explain the goal is here: // Start of script; Names Default To Here( 1 ); Clear Log(); dt = Open( "$SAMPLE_DATA/Bands Data.JMP" ); Graph Builder( Size( 534, 456 ), Show Control Panel( 0 ), Variables( Y( :viscosity, Combine( "Parallel Merged" ) ), Y( :proof cut, Position( 1 ), Combine( "Parallel Merged" ) ), Y( :blade pressure, Position( 1 ), Combine( "Parallel Merged" ) ), Color( :viscosity ) ), Elements( Points( Y( 1 ), Y( 2 ), Y( 3 ), Legend( 10 ) ) ) )

FN · ‎07-22-2020

Within the marginal model plots, one can activate the option to see the data. The main problem is that each plot is individual, so if you want to change dot size or transparency you need to do it one by one.

FN · ‎07-19-2020

The profiler can optimize the desirability. Is this interface available as well? Would it solve NLP or MILP problems?

FN · ‎07-19-2020

This is great! Thanks for sharing* it Regarding Python in Windows, Anaconda with Python 3.6 works out-of-the-box with JMP 15, but you need to update all the packages since it is a 2 y old package. I hope JMP improves their Python support in next versions. (*note) The interface you have built can be used for other python APIs. Have you published the source code as well? Regards.

FN · ‎07-18-2020

I really like the new feature in JMP 15 showing the distributions below the headers. Yet, for time series (sensor data), distributions are not enough. It will be great if we can also add spark-lines (Excel and many other software have them).

FN · ‎07-10-2020

Batch processes have different phases (e.g. loading, waiting, pre-heating, reaction, cooling, curing, unloading). Dynamic time warping can be very tricky if you try to align all those phases at once. The alternative is not pleasant, as you have to create a model per phase. As a record of these phases is available (in a separate column), it should be easier to pass this information to the FDE platform so the dynamic time warping algorithm knows what pieces need to be aligned.

FN · ‎07-10-2020

Gaussian Processes are a very powerful modeling technique in predictive analytics. https://scikit-learn.org/stable/modules/gaussian_process.html While JMP has GP, it is only for specialized modeling (not possible to fit using training/validation/test)

FN · ‎07-10-2020

I believe that a better integration with Anaconda is really necessary. Today, the only supported version is Anaconda with Python 3.6 (a 2-y old package) Alternatively, the ideal scenario will be to have an option in settings for specifying the PATH. This way, portable versions of Python/R can be used. See for example: https://winpython.github.io/

FN · ‎06-09-2020

If you have JMP 15 (2019 release), you can do it now https://www.jmp.com/support/help/en/15.1/#page/jmp/import-data-from-google-sheets.shtml

FN · ‎06-09-2020

I know it is an old discussion, but JMP 15 includes this option now. https://www.jmp.com/support/help/en/15.1/#page/jmp/import-data-from-google-sheets.shtml

Date Registered	‎01-26-2019 05:36 PM
Date Last Visited	Thursday
Total Messages Posted	91
Total Kudos Received	26

Online Status	Offline
Date Last Visited	Thursday