Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Re: Combine Multiple Single Punch/Binary Variables into One Variable

Topic Options

- Start Article
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Dec 1, 2014 8:00 AM
(4504 views)

I have a set of data where we asked in which industry a respondent works. They are given a small list of options and can only select one. When the data comes back from processing, that single question is blown out to one variable per option rather than a single variable with multiple coded labels. What I am trying to figure out is how to use a JMP formula to read each of the variables within a case and apply the appropriate coded label in a new variable based on where the 1 or "Yes" is located.

I could easily do this in Excel and import the data back into JMP, but I would really like to learn how to do this within JMP.

Does anyone have a good example of collapsing multiple binary variables into one using formulas?

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

So what I ended up doing is creating a nested IF shown below. I then went in and added Value Labels for each of the points and I get essentially what I am looking for but I am curious if there is a better way to get where I have ended up. I have attached the sample data which includes my solution.

Couple notes on what I wrote, the final if statement is completely unnecessary and having Error outside of quote would, ironically, cause an error.

8 REPLIES 8

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

I attached a sample to my original post.

Respondents are asked a question like...

In which of the following industries do you work? (Select one response)

Industry 1

Industry 2

Industry 3

Industry 4

Industry 5

Industry 6

Industry 7

Industry 8

Since the respondent can only select one response, I would expect this question to be output as a single variable with each of the possible industries as a coded option for that variable. What I get is a variable for each option.

Does that, with the sample data, make what I am looking for more clear?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

Here's one way:

1. Create a new column, for example **Response**, and make it **Character** type

2. Click on **Column Properties > Formula**

3. Click on **Edit Formula**

3. Double click on the formula rectangle and enter the following code:

If(:worksc1, "Worksc1", :worksc2, "Worksc2", :worksc3, "Worksc3", :worksc4, "Worksc4", :worksc5, "Worksc5", :worksc6, "Worksc6", :worksc7, "Worksc7", :worksc8, "Worksc8")

Basically it's a giant IF statement that looks for Yes values (which are coded to 1s). The Response column will now show which column contained Yes.

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

Thank you PMroz, it looks like I ended up where you did as you can see in my post below.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

The : prefix tells JMP that this is a column. You can only use :**column name** if there are no "funny" characters in the column name, like /&*^%$# etc.

JMP will evaluate any positive integer to TRUE, so I just used the column value instead of comparing it to 1.

These two sets of JSL statements are equivalent:

abc = **1**;

if **(**abc, print**(**"abc is True"**))**;

xyz = **1**;

if **(**xyz == **1**, print**(**"xyz is True"**))**;

The log shows this output:

**"abc is True"**

**"xyz is True"**

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

Thank you for clarifying the logic in your code block.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

So what I ended up doing is creating a nested IF shown below. I then went in and added Value Labels for each of the points and I get essentially what I am looking for but I am curious if there is a better way to get where I have ended up. I have attached the sample data which includes my solution.

Couple notes on what I wrote, the final if statement is completely unnecessary and having Error outside of quote would, ironically, cause an error.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Combine Multiple Single Punch/Binary Variables into One Variable

Another way: Stack your data table (*Tables menu*) and delete all No rows (Right-click on a "No", pick "Select Matching Cells" and then Delete Rows from *Rows menu*).

Same in JSL:

dt = Data Table**(** "Data Subset.jmp" **)** << **Stack****(**

columns**(**

:worksc1,

:worksc2,

:worksc3,

:worksc4,

:worksc5,

:worksc6,

:worksc7,

:worksc8

**)**,

Source Label Column**(** "Label" **)**,

Stacked Data Column**(** "Data" **)**

**)**;

dt << **select where****(** :Data == **0** **)** << **delete rows**;