Have an idea?

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

How to remove the selected levels from attribute

I am facing some problem about removing the selected levels from attribute which will be used in ACBC

I am using RIC(), but seems it does not work as I got error message "Sawtooth Error # 174 ACBC error.
Please check the error log for details.
Please try to refresh your browser or try to backup and submit again. If the error continues please contact the survey administrator."

is there any ways that can remove the selected levels from specific attribute and the remaining levels  as the levels of the attribute in ACBC?
asked Nov 25 by Sangho (120 points)
It looks like that error is meant to show up when one of the level lists is completely empty.  Is it possible that that is what is happening in your case?  Can you share your full constructed list instructions with us?
please check what i did below
RIC only removes items from a list.  You need to use ADD or a similar list operation to first add items to the list, then call RIC to remove the ones you want to get rid off.

1 Answer

0 votes
I simply created four constructed lists(: BrandList, CapacityList, DropBrand, DropCapcity)

###predefined Attrbitue(: Brand, Capacity)###

[Brand]
1. A
2. B
3. C
4. D
5. E

[Capacity]
1. 1000
2. 2000
3. 3000
4. 4000
5. 5000

###Constructed list###
[BrandList]
Add(Brand)

[CapacityList]
Add(Capacity)

[DropBrand]
RIC(Q1)

[DropCapacity]
RIC(Q2)

after done making constructed list above, I have set DropBrand in List of Levels for Brand and Drop Capacity in List of Levels for Capacity in Attributes tab in ACBC Exercise Setting.

I thought that this logic would filter the selected levels through Q1 and Q2(which are the multiple selection) and remain the unselected levels as levels of Attributes,Brand and Capacity in ACBC.
answered Nov 26 by anonymous
It doesn't seem like you need 4 constructed lists.  Don't you want one constructed list with

Add(Brand)
RIC(Q1)

and then a CapacityList with

Add(Capacity)
RIC(Q2)
Thank you for your comment.

I have further question below:

1. if we use ANC, it add the unselected levels of a attribute into ACBC.  
in this case,  what utility sawtooth provide us for unselected level of attribute by a respondent by using HB utility estimation?

2. through ANC, if more than 7 levels of attributes are put into ACBC, then how utility value are unstable and does HB utility estimation work, even if there is more than 7 levels of a attribute included in the estimation?

Thank you
Keep in mind that constructed lists are kind of separate from the utility model.  If you have 7 brands, the utility model for each respondent will always have 7 parameters.  If you build a constructed list and a respondent only filters in 5 brands, they will still have 7 parameters in their model.

See https://www.sawtoothsoftware.com/help/lighthouse-studio/manual/index.html?customized_constructed_attri.html for additional details.

The best rule of thumb is that you want to make sure the non-BYO levels of an attribute are seen 2, preferably 3 times by a respondent.  So it's not really that you can't go above 7 levels in an attribute, it's just that you either need to create a larger pool of screener concepts (more screener questions),  increase the amount of variation from the BYO, or drop the BYO question so each level is used equally in the screener section.  

HB estimation will always "work," even if a respondent never sees one of your brands.  It will infer from the sample as a whole (basically an educated guess), which seems pretty sub-optimal if you are trying to build individual-level models.  Ideally it makes sense that you want a survey respondent to see every level of an attribute several times.
...