Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

How can I analyze CBC data if I have only fixed Choice tasks?

I encountered the same problem. My design contains 18 fixed tasks. When I want to analyze the data I always receive the Error message "There are no respondents with CBC data to analyze." Is there a way how to fix this problem?

Thanks in advance for your help!
asked Jun 9, 2017 by jane
Thank you so much for your help Brian!! It worked now.

1 Answer

0 votes
Best answer
I hope your fixed design was created in a way that will support efficient utility estimation...I'm assuming so for the following suggestions.

One way to fool the software into doing this is to make a new study in Lighthouse Studio for the purposes of analyzing your CBC data.   Enter your attributes and levels again exactly as you had them in the original study.  And, the same number of choice tasks, number of concepts per task, etc.

In that new Lighthouse project, generate a new design (say, balanced overlap) with as many versions as your original fixed design (presumably 1).  From the Design tab, click the Export button to export the design to a .CSV.  Replace the design Lighthouse just generated with your design from the original study (you can export that original design to a .CSV file so you can just paste the info out of it into the new .CSV file).  Import the modified .CSV design file into the new Lighthouse project (using the Import button from the Design tab), so now you have your fixed design represented within "Random" CBC tasks.

From the original project, export the responses to the original fixed tasks to a .CSV file.  Format these for paper-and-pencil import with the new project.  In the new project, use Field + Create Accumulate Data Template File.  That gives you a .CSV with just header rows in it.  That's the starting point for pasting the raw respondent answers to the choice tasks into.  In the new project, use Field + Accumulate Paper & Pencil Data to read the "paper and pencil" data from the .CSV file into the new Lighthouse project.  That should be it.  You may need to copy other segmentation data from your original study into this new one if you need them for segmentation, covariates, or weights.

Perhaps somebody else has a better trick for doing this, but it's an approach that will do it.
answered Jun 9, 2017 by Bryan Orme Platinum Sawtooth Software, Inc. (153,180 points)
Hi Brian

Thanks for your response. I tried to fool the software in the way you described but it does not work. I assume that I am making a mistake though.

My current design looks like this:
"Version (0 = Fixed Task)","Task","Concept","Att 1 - source","Att 2 - location","Att 3 - owner","Att 4 - impact","Att 5 - costs"

 In this original design I have 18 fixed tasks and 2 concepts per task, none option = traditional. All respondents will receive a questionnaire with the same choice sets.

What I did now is that I created a new project where I entered the attributes and levels in the same way as in my original design. Then I created a design an entered:
- Random Tasks: 18
- Fixed Tasks: 0
- Concepts per Task: 2
- None option: traditional

I exported the design and received a file with 10801 rows. I think now I am unsure what to replace with what in the new-design-csv file. Shall I delete the 10800 rows and just leave the header of the new design and copy-paste the 36 rows of my original design into the new design?

When I do it like this and import the design into the new study I receive an error message "The end of the file was reached before reading a complete design. The file should have 0 fixed task rows (if any) and 10800 random task rows (if any).

Or shall I create the new design in a way that I have 17 fixed tasks and 1 random task?
In the new CBC Lighthouse project, you need to tell it on the Design tab that there is just one version of the questionnaire with 18 tasks.  So, your version # will be "1" for all 18 tasks (not 0).  And, you'll have just the 18 tasks listed in the design file.

By the way, version="0" refers to fixed tasks in the design.  So, you should start listing your "random" tasks as version="1".  And, you just have 1 version, it appears.
I'm encountering the same problem. I followed the steps as described, but I also get the same error ""The end of the file was reached before reading a complete design. The file should have 0 fixed task rows (if any) and 10800 random task rows (if any)."

As you commented, I changed the version from "0" to "1", to all tasks, but I still get the same error. Is it possible to send my files to you so you can look at it? I am struggling with this for some days now and I don't know what I can do more to get this problem solved.

Thanks in advance.
The error message you are getting is saying that the software is expecting 10800 rows in the design.  Since you have 18 tasks and 2 concepts per task, that means the software is expecting you to be providing 300 versions.  300 versions is the default in the software.  So, it seems to me that you have not changed the number of versions from the default (300) to the number of versions you are trying to read in.  You should be telling it that you have just 1 version in your design, not 300.  Then, try importing your .CSV file again.
Thank you for you comment, Bryan. The trick worked and I was successfully able to import the fixed tasks design as random tasks now. When I followed your further instructions step by step, and I click on analysis again, I get the following error message: "The column does not exist in the data file. It is a required column". I looks like it didn't import the CBC files in a correct way?
It looks like it (partly) worked out. I can analyze my CBC data now. But lastly, I need to import all the segmentation data as well. I copied them into the RAWexport excel file and imported them into my new conjoint dataset (Field + Accumulate Paper & Pencil Data), like I did with the CBC data set. But how can I copy all the segmentation questions into the new file? Do I need to draft them again one by one, or is there a quicker and easier way? Also, how can I be sure that the segmentation data will be matched exactly with completely with the questions?
Under File + Data Management...   go to the Get Data tab and then use the "Merge External Data" option there.  Merge the extra data from a .CSV file containing matching respondent numbers and the additional fields.

Or, you can just go to the "View/Edit Data" tab and click the "Add Column" icon at the right of the dialog to add new columns to the data file.  Then (making sure your respondent numbers match up with the data for each row) you can paste in the additional variables you need to perform analysis.