Dear all,
I read your article "Consistency Cutoffs to Identify "Bad" Respondents in CBC, ACBC, and MaxDiff" and I understand when having a CBC we should not elminate respondents based on low RLH values alone - but also look at other variables like time & straightlining behavior.
I have done what you recommended in the above mentioned article. I generated random responses and looked at the median RLH (0.32) and the 95 % percentile (0.43) for the generated responses.
Now, I have real respondents in my sample who have RLH below 0.4 or even below 0.3. but I can not see any suspicious behavior to identify them as "bad respondents".
Should I delete them or keep them?
Thanks a lot!