3

I just did a survey and I'm pretty sure some people just filled in the bubbles on the survey without reading the questions. A pretty obvious indicator are responses to items that are negatively coded (e.g. answering "strongly agree" to related items like "I'm a moron" and "I am not a moron"). Are there any methodological guidelines for finding and removing these responses? My results change quite a bit depending on my method for filtering people out. I've been toying with the idea of removing people whose standard deviation between all survey items is below a certain threshold. But, I'm unaware of validated guidelines for removing participants.

Jim
  • 261
  • I don't know of any such guidelines - except "use your judgment" I'll be interested in what answers you get! – Peter Flom May 08 '13 at 23:29
  • You can treat it as non-response ("My hands clicked on the bubble, but my brain was elsewhere"). You can delete them safely if they are MCAR (morons completely at random), but be prepared to find out that being an idiot is correlated with the variables of interest... – Deer Hunter May 09 '13 at 18:47
  • 1
    Related question on another site: http://cogsci.stackexchange.com/questions/3267/analyzing-control-questions-data-for-a-survey – Gala May 16 '13 at 18:31

0 Answers0