NIPO ODIN Version 5.17
How does filtering in NIPO Conjoint Measurement actually work?For each compare test between a set of two profiles, the system first picks a number of levels randomly, then selects two profiles out of them that are utility balanced, then looks whether the combination of levels already have been presented, and then tries if it is allowed according to the filter conditions. If a combination has already been presented, or the profiles are not compliant with the filter this procedure is repeated. After 25 tries the combination-repeat check is ignored, after 50 tries the filter is ignored. This implies that if the filter is too demanding, you might end up that the filter will be ignored totally and / or two profiles might be presented twice. The advantage of this procedure is the random pick. However, in combination with the complex filter possibilities it makes the process unpredictable for the researcher. Question is whether this "search" function still works according to the "specifications" or users expectations. In practice the utility balance function supports the filter function indirectly. The filter function is aimed to block unrealistic profiles and most of the time this function will produce such realistic profiles. In tests that are not aimed to mimic these respondents, but merely to check the filter function, this support will not be noticeable. |
|||
|
