ApolloChoiceModelling forum

Hi there,

I am trying to clean my data at the moment.
So far, I´ve looked for straighlining within the choice tasks and other survey questions and also for speeders who spend less than 3 seconds on one choice task. Furthermore, I took a look at suspicious answers to open qualitative questions in the survey.
Is that approach sufficient or would you recommend any other analysis to find "bad respondents"? Also hints on helpful literature on this topic are very welcome.

Thank you very much in advance,
Best regards,
Patrick

Patrick

removing people arbitrarily according to straightlining or speeding is bad practice. These people might still be behaving rationally. There is ample literature guidance on this. Much better to look at why someone might be behaving in a certain way, and see how you can accommodate them in a model

Stephane

Dear Stephane,

thank you for this important advice.
Though, it might be the same with irrational respondents (not answering two same fixed choice tasks in the same way), isn´t it?
So not deleting them from the sample just because they failed the test for completeness axiom?

Thank you very much.

The key job for the modeller here is to judge whether the behaviour is in line with RUM or not

Are there any specific guidelines how this can be judged or any papers you can recommend on dealing with inconsistent respondents?

Patrick

a long time ago, I wrote this paper: https://doi.org/10.1016/j.trd.2010.04.008

There are probably many others now

This is in terms of "inconsistent" responses. For people that are very fast or very slow, my recommendation is always to first see whether these people actually behave differently from others

Stephane