Chrissl11 wrote: Don't know where my mind is at the moment...

Is there a reason why you choose LimeSurvey to do a choice based conjoint (CBC)?
LimeSurvey isn't offering any help (visually and conceptional) when it comes to CBC.

When doing it in LimeSurvey (without workarounds) the None Option is not shown side by side.
The answer buttons are not directly shown under the concepts (3 concepts + Noneoption).

To implement the experimental design is a lot of work:
How many attributes do you have?
And what is maximum of levels the attributes have?

You mentioned 50 test groups. What do you mean with test group? A task (question with 3 concepts and a none-option)?

What kind of tool do you use to analyze the choice data?


Personally I wouldn't change the images in the task. I would create a dedicated question(group) for every task. When it comes to debugging, your approach can drive you mad. You might have a little mistype in one concept and won't find the issue. If you have one task per question(group) you can print them out, do screenshots with a fixed question name.
In Conjoint is no real random factor. You create as many testplans as possible (today you often use 1000-2000 ones, to ensure you don't have framing effects or order effects) and then try to distribute them over your respondents. An easy way to distribute is to use a random number which chooses the testplan.

@tpartner:
Thank you tpartner for your solution!

As far as I checked it, it is working.
And please appoligize, I did not know, that it is necessary to provide only the relevant questions as jelo said "You can just create a small survey and attach the export here. You don't need to provide the complete survey." so I thought it is up to me and of course, I've chose the easiest way. But I will keep that in mind!

But what do you mean by "Minor - when generating a random number, you should always put in a check to see if the number has been previously generated". I thought actually when having 1000 participants, that (almost, because it is random) every 50st will start with no. 1 again. Is that a wrong assumption?



@jelo:
There are several reasons!
First, the panel provider I will be using, uses LimeSurvey too. Therefore, it will be very easy by simply providing them the final file. Second, I can use the Sawtooth company account of a good friend to analyze the data. But I cannot carry out my survey on that, because it is her company account. I checked a lot of different providers besides Sawtooth, however, they are all really expensive - above 1,000 euros per licence. That is why I have chosen LimeSurvey: It provides a lot of possibilites and a proper support.

Yours,
Chrissl

Chrissl11 wrote: I thought actually when having 1000 participants, that (almost, because it is random) every 50st will start with no. 1 again. Is that a wrong assumption?

Tony meant the issue, that the random number in a equation changes every time it triggered (e.g. a respondent is switching back and forth in a survey). So the random number would change during an interview and the respondent would get different questions when going forwards again.

That is prevented via {if(is_empty(random), rand(1, , random)}. The random number will only be assigned once.

The random number is random. So you don't get a 100 percent correct distribution between 1 and 8.
If you have enough respondents you will get a proper distribution.

Chrissl11 wrote: First, the panel provider I will be using, uses LimeSurvey too. Therefore, it will be very easy by simply providing them the final file.

Panelprovider runs the survey with LimeSurvey? Interesting. Splendid?