For client reasons, I'm not able to use partial profile CBC or ACBC. I will also be exporting the design from Lighthouse Studio and sending it to an external vendor for programming/hosting outside of Lighthouse Studio.
My conjoint has 15 attributes - Brand, 13 feature attributes, and Price. I would like each conjoint screen to show 7 attributes: Brand, 5 feature attributes, and Price. Instead of partial profile CBC or ACBC, I would like to generate a full profile CBC design with all 15 attributes where each concept (in the design file) shows all 15 attributes. I would like to then instruct programmers to hide 8 feature attributes on screens for each respondent when they complete their conjoint (these attributes will be based on a question prior to CBC where respondents choose their 5 most important attributes - these will be shown in each respondent's conjoint). So each respondent will see 7 attributes in their conjoint screens - even though the design for each respondent has 15 attributes.
I would like to analyze the data via default HB in Lighthouse to obtain respondent-level utilities. For analysis, I was planning to read in the conjoint response data as usual for CBC paper-and-pencil via Lighthouse Studio's template. I would then run default HB analysis to produce utilities for each respondent for all 15 attributes/all 15 attribute's levels - based on HB's ability to "borrow" information across respondents. The assumption is that attributes not seen by respondents will be filled in/imputed by HB. So the design/analysis assumes all 15 attributes seen by all respondents - even though this wasn't the case. Is this OK to do OR do I need to use a modified process for Lighthouse HB analysis - if modified, do any step by step instructions exist for this for someone who hasn't done it before?
Also, do I need to be concerned that Brand and Price will be shown to all respondents and the remaining 13 attributes will be shown to varying numbers of respondents - with some attributes potentially being shown to very few respondents if the feature attribute is unpopular amongst respondents?
Thank you.