choice-metrics.com

Dear all,

I am trying to block an existing orthogonal design (http://www2.research.att.com/~njas/oadir/oa.32.9.4.2.txt) with Ngene. I deleted the 9th column of the existing design (I need only 2x4 attributes) and used the following syntax:

Code: Select all

? This will block an orthogonal design with 32 choice situations into 4 blocks.

Design
;alts = scen1*, scen2*, none
;rows = 32
;block = 4
;eval = oa.32.9.4.2.xlsx

;model:
U(scen1) =
b1          +
b2 * forest[0,1,2,3]  +     
b3 * settle[0,1,2,3]  +
b4 * flood[0,1,2,3]   +
b5 * cost[0,1,2,3]
/     
U(scen2) =
b1          +
b2 * forest[0,1,2,3]  +
b3 * settle[0,1,2,3]  +
b4 * flood[0,1,2,3]   +
b5 * cost[0,1,2,3] 
$


However, instead of just blocking the existing design, Ngene creates a completely new, blocked design. Can anyone tell me what I am doing wrong?

Alternatively: could I use the 9th, skipped column to block the design into 4 blocks? I remember reading that somewhere, but I am not sure whether it is actually working.

Cheers,
Andrea

Hi Andrea

Ngene requires the design to be presented in a very particular data format which is explained on P21 of the manual. I will describe this format below.

The program requires

1. Each column to have a heading name.
2. The first column should be a series of numbers correspond to respondents. In your case, you only have one respondent (design) so it should just be a vector of 1s. This column is used by the program to associate a design with a particular respondent or segment. This relates to the more advanced feature of the program where you can generate designs which is described in S8.4 of the manual. The heading title design not matter - that is, you can use any name you want.

We understand that this first column is a little bit of an annoyance, but it is necessary for the more advanced features to work unfortunately.

3. The second column should represent the choice task number. The first cell should be a 1, the second cell 2, up to the number of choice tasks in your design (32) in your case. The heading title design not matter - that is, you can use any name you want.
4. The remainder of the data are the attributes and the levels.

When I set your data up using this format, the program did exactly what you want (i.e., keep the design and generate a new blocking column.

I hope this helps.

Hi John

Thanks for your help, now it works perfectly! I did use the right data format (without headers though), but did not uncheck "with column headers" in the session options. Beginner's mistake...

Cheers,
Andrea