Have an idea?

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

LC utilities in online simulator


I would like to use the utilities I received from Sawtooth Latent Class module in the online simulator. Is this possible?

I have difficulties uploading these utilities. I have a .csv file with the utilities, but I did not receive a .labels or .xml file with the labels from Sawtooth.

Can I find this file somewhere? Or can I make it myself?

asked Apr 3, 2012 by anonymous

1 Answer

+1 vote
First, you need to get the utilities out of SMRT into the right format.  Within SMRT, go to Run Manager, select Export, and under "Save as Type" select "Generic Conjoint File (*.hbu)".  Try uploading that utility file (*.hbu format).

SMRT converts the group-based utility information into pseudo individual-level utilities.  If you are interested regarding the math for that, it is described in the SMRT manual, under the topic "How the simulator uses latent class data" (it's one of the Appendices in the SMRT manual).

Essentially, we create a weighted average of the utilities from the groups for each respondent, where the weights are the probabilities of membership that each respondent belongs in each group.  So, if I am 60% likely to belong to group 1 and 40% likely to belong to group 2, then my individual-level utilities are given by taking .6 times the utilities from group 1 + .4 times the utilities for group 2.

Pseudo individual-level Latent Class utilities developed in this way are generally considered inferior to HB utilities as developed under the CBC/HB system.
answered Apr 3, 2012 by Bryan Orme Platinum Sawtooth Software, Inc. (187,915 points)
Thanks for your quick reply Bryan Orme!

I cannot use the SMRT program as my university has no license for it (only for SSI Web, HB and LC) (University of Groningen, the Netherlands). This analysis is for my master thesis.

I used the .hbu output of the HB program as input in the online simulator. This worked fine. I received share of preference's in the online similar that were equal to the one I calculated myself in excel. However, apparently I needed to use a different .hbu output?

I already calculated these pseudo individual-level utilities in excel to be able to calculate the hit rate of the model. So in theory, I would be able to upload these to the online simulator. Can i save them in excel as .csv and upload this in the online simulator? However, again, there is the .label/ .xml file for the labels I do not have....

I already calculated share of preference's in excel and did sensitivity analysis. However, I would like to use the online simulator because of the Randomized First Choice rule, which seems like to much work to compute in excel.

Hope you can give me more information on how to proceed. Thanks!
SMRT is a free license to people who are licensed to use CBC, ACBC, ACA, or CVA within the SSI Web platform.  But, now I understand that your Latent Class utilities came from the standalone Latent Class system.

Now that you have pseudo individual-level utilities (that you computed yourself), why not just paste those into a .hbu format, following the .hbu file layout that you got from the CBC/HB output?  Then, upload that to the Online Simulator?
That seems like a perfect solution! However, it is not clear to me how I can open and save a .hbu file. It seems like I cannot save a .hbu file in excel.

Would't it be easier to use a .csv file to upload to the online simulator, as I can save the utilities in an excel file as .csv? However, how would I obtain the .labels file?

Thank you for all your help!
The .hbu file is simply a text file that you can open and modify with a text editor like Notepad.  You'll find that there is a header section at the top that describes the layout of the file.  But, soon you'll see respondent numbers followed by a bunch of numbers that include utility values.  You can simply swap in your Latent Class utilities and resave the .hbu file as a new file.  The .hbu file is something I have used for years, so that's familiar to me.  I'm not very up-to-speed on the new .csv file layout and labels file.  I would suggest you call our technical support line for more information about those files and layouts.