Have an idea?

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

Did the MBC model actually use covariates even if dataset didn't have the variable?

Model indicates covariates were used but the variable specified doesn't exist in the dataset.  

I initially estimated a model using a covariate and subsequently decided not to use the covariate.  I removed the variable originally used as a covariate from the dataset and pulled in the new dataset.  Before modeling the new data, I forgot to unclick the "Use Covariates" check box so the covariate variable was still specified when running the new model.  Did the modeling actually take that into account even though the variable it was looking to use was not in the dataset?
asked Jan 20, 2014 by anonymous

1 Answer

+1 vote
If no variables in the list of covariates are checked to be included, and yet the software is selected to use covariates, then a standard alpha will be used.

If you changed your demographics file to remove the variable, and did not refresh the list of covariates (the prior checked covariate remains in place), then the software would take the column that the previous variable occupied as the covariate column.

If you change the demographics file to a different one within the software, then the software should notice the change and automatically update the list of covariates.  However, if you edit the existing file outside of the software, then it will not recognize the change.
answered Jan 20, 2014 by Walter Williams Gold Sawtooth Software, Inc. (15,275 points)
Thanks Walter.  I removed the demographics file completely and opened a new dataset that did not include the variable that was initial used as a covariate.  Can I assume that a standard alpha will be used in this case?
By dataset do you mean the respondent data file, the demographics file or a new project?
By dataset I mean data file - at the task level.  I removed the demographics file from the 'Data Files' tab.  The project is the same project as I used when I started to model the model with covariates but then subsequently cancelled that model.
I believe in that case the standard alpha will be used.