This question is a little odd because you are asking about a data collection method (stated choice survey) vs a statistical method (PCA). PCA is a stat method that often is applied to a group of potentially related survey items (eg, the items on an instrument used to assess health status, life satisfaction, knowledge, attitudes, or behavior). It clusters those items into a smaller number of reasonably independent components. So for example, in health care, the widely used Short Form 12 health assessment clusters 12 items into a physical health component and a mental health component. And yes, one certainly could try valuing those PCA components using a stated preference survey. The strong finding in recent literature is that the traditional stated preference survey on willingness to pay is inferior to a discrete-choice-experiment survey approach. In the DCE survey, the dimensions that vary in level could be determined by a PCA on prior attitudinal survey data or by the researcher's judgement AND -- more importantly-- by the information needed (e.g., a road-building investment budgeter wants to understand preferences between saving more travel time or sacrificing travel time to increase road safety, determining the balance the public desires between these two competing demands). You could do a survey that includes both stated preference and DCE questions aimed at valuing the same attributes. But we already have shown that the DCE survey gives you cleaner, more reproducible, lower variance estimates. Better to invest that same survey time in an additional DCE question set that broadens the range of tradeoffs you've assessed. The main quality constraint in a DCE is that the number of levels one can ask about for any given attribute both is small and is defined by the survey designer. The worst DCE mistakes result when the DCE choice range is not thought out. (e.g., in the experiment above, in n early DCE survey, because the choices were not thought out carefully, they had too high a floor. They essentially equated to an hour of travel time is worth half as much as an hour of life, an equal amount as an hour of life, twice as much, four times as much, or 8 times as much. Needless to say, wise researchers should ignore the published results of that survey.) That second DCE question gives you a choice to broaden the range over which your attribute tradeoffs focus. That's a far better use of survey time than adding a SP question.

Thanks for your answer!

The reason for using two survey methods is that I think that finding the importance of some measurable attributes (like time, cost) is more accurate with SP survey than some attitudinal questions. That's why I am trying to use two surveys, measurable attributes like cost and time with SP survey and attitudinal attributes like safety and comfort with PCA method.

But to create a link between two methods and to find the relative importance of attributes in two surveys (travel time from SP and safety from PCA) I want to put one common attribute (say comfort) into both of them to make a connection between all variables.

So, my SP attributes are time, cost and comfort and my probable clusters in PCA will also comfort, safety and quality. So, I can find the relative importance between "safety" and "time" by having "comfort" into both models.