Marking an attribute as Continuous lets you more intuitively define product alternatives and more easily conduct linear interpolation between attribute levels within market simulations.
Some attributes like speed, price, and weight are quantitative/continuous in nature, such as 30 kilos, 40 kilos, and 50 kilos. It is common to simulate the likely preference for products defined between the levels of such attributes using the market simulator. (We use linear interpolation to estimate the likely preference for amounts between levels included in the conjoint experiment.)
Continuing with the weight example above, if you specify that values of 30, 40, and 50 are associated with levels 1, 2, and 3; then to define a product in the simulator that weighs 40 pounds, you simply type a "40" as the product specification. If you want to specify a product that is 36 pounds, then you just type a "36".
There are two main reasons to allow "Not Applicable" (N/A) to be used to specify the level for a product in market simulations:
1.To indicate for an alternative-specific design (currently not supported in Discover) that an attribute does not apply to the product concept. For example, advanced conjoint designs can involve alternative-specific attributes, such as "frequency of arrival" only applying to buses and "parking fee" only applying to cars; though both buses and cars are ways to get to work. Thus, you would specify that buses are N/A on the parking fee attribute and cars are N/A on the frequency of arrival attribute.
2.To turn off the effect of an attribute across ALL products in the scenario: Specifying that all products have N/A for an attribute adds zero utility for this attribute to all products in the market simulation. (Note: since conjoint utilities are usually zero-centered, adding an attribute as N/A is the equivalent of using the utility for a level of average preference for that attribute. Thus, specifying N/A is not the same as saying that a product doesn’t have a level of a given attribute. If an attribute doesn’t have a level of an attribute, then a null or blank level should have been included in your original conjoint design so that a utility weight was captured to reflect relative preference when a product lacks this feature.)
When Randomized First Choice is used, we recommend turning off the correction for product similarity with respect to the Price attribute. Checking the box in this table accomplishes this, by informing the Randomized First Choice algorithm to apply independent rather than correlated error to the price utilities when multiple products share the same price level.
Failure to identify the price attribute so that Randomized First Choice can avoid correcting for product similarity for price can result in strange and abrupt "kinks" in derived price curves via sensitivity analysis. Even for standard market simulations involving three or more product alternatives, failure to inform Randomized First Choice of the price attribute can lead to overprediction of share of preference when a product is unique with respect to price and many other products lump together at the same price.