Deriving one attribute from another for an alternative

[neeraj85]

Hi,

Consider the following case:

I have defined an alternative (ALT1) by using two attributes Travel time (TT) and Time spent in congestion (Tcong). The levels for the attributes are:
TT[0.8, 0.9, 1.0, 1.1, 1.2]
Tcong[0.3, 0.4, 0.5, 0.6, 0.7]

These levels represent the multipliers which will be multiplied with the values revealed by the respondent.

E.g. If respondent gives --> TT = 20 min. Then I wish to do the following:

ALT1
TT = 0.8*20 = 16
TCong = 0.5*16 = 5 (I multiplied the calculated TT value further with one of the levels of Tcong)


Will this operation have a negative impact during the model estimation (as the level for Tcong is effectively 0.8X0.5=0.4)?


Please advise whether is it a safe operation or should I refrain from doing this.


Thanks
Neeraj

[johnr]

Hi Neeraj

This sounds like a (more sophisticated) form of pivot design. See this article for a discussion of this:

Rose, J.M., Bliemer, M.C.J., Hensher, D.A. and Collins, A.T. (2008) Designing Efficient Stated Choice Experiments Involving Respondent Based Reference Alternatives, Transportation Research Part B, 42(4), 395-406.

I would suggest you simulate a data set and see what happens in this case, given that you are pivoting two attributes from a common value. I suspect you are fine given that each attribute has different levels meaning that they are not perfectly correlated, however simulations should confirm this is the case.

John

[neeraj85]

Thanks a lot sir.

Can you please explain what do you mean by simulate a dataset?

Does it mean analyzing the pilot survey data for any correlation between TT and Tcong?

Is there a way I can simulate a dataset using Ngene or some other software package?

Thanks
Neeraj

[Michiel Bliemer]

John mentions pivot designs and perhaps they are suitable.

Ngene can create relative designs for you. You first need to specify segments with reference levels (such as short distance, tt = 15 min, tcong = 5 min, and long distance, tt = 45 min, tcong = 15 min), and then you need to specify pivots. If you want multiplications you can specify percentages.

For example, you can specify tt with levels [-20%,0,50%], which multiplies a given base level (say tt = 20) with 0.8 and 1.5 to create the actual levels. For more information I refer to Section 8.2 of the Ngene manual.

[neeraj85]

Thanks a lot.

I went through the section 8.3 of the manual on Pivot designs.

The requirement for my design resembles to the one mentioned in a previous thread "Pivoted design without status quo alternative" by Arnaud Blaser on Jun 27, 2014.

I don't wish to display the status quo alternative and just want to use its info to calculate the travel time for both alternatives A and B.

Eg. if responden'ts TT = 30 mins; TT_A = 0.8*30=24 mins and TT_B=1.2*30=36 mins ... Where 0.8, ..., 1.2 are the attribute levels for travel times for both the alternatives.


Can we accomplish this objective in ngene now?

Thanks
Neeraj

[Michiel Bliemer]

No, one needs to specify the reference levels in one of the alternatives. It is on our wish list to make this more flexible in the future.