First, you must create a design. In your example, you listed 4 binary attributes. That means that there are 16 possible combinations. Your design could be a full factorial design with all 16 possible combinations. Take a look at the ORTHOPLAN documentation.
Next, you must collect data from multiple respondents. You could create
"cards" showing each possible product and ask the respondent to rank or rate the cards. More researchers use ratings than use ranks. Take a look at PLANCARDS.
Once you have the data, you analyze them using the CONJOINT procedure.
CONJOINT essentially does a within-persons OLS regression. What you can
learn from the CONJOINT procedure are attribute importances and the relative preferences for the levels of the attributes.
Lastly, you can use CONJOINT to do so-called simulation. This is where you
can model share of preference for possible products with combinations of
attribute levels.
There are many issues in the use of conjoint analysis. As the number of
attributes and levels goes up, it becomes unreasonable to use a full profile design, for the resulting rank or rating task would prove tiring for the respondent and most likely produce data with serious validity and reliability problems. Therefore, researchers turn to fractional factorials and other designs. Do you expect interactions between attributes, or will a main effects design suffice? Also, you want to create a data elicitation method that simulates the real decision process, meaning you might use physical props or multimedia. Regarding traditional conjoint analysis as a method,
some researchers object to the method itself for the reason that eliciting
preferences does not tell much about what respondents would actually choose in a buying situation. For this reason, some researchers prefer choice analysis to traditional conjoint analysis.
Anthony Babinec
发帖
雷达卡
京公网安备 11010802022788号
