Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Minimum Fit Statistic to Identify Random Responders with 99% Correct Classification?

Just a quick question. When running the simulation to come up with the 95% correct classification bounds (https://www.sawtoothsoftware.com/help/issues/ssiweb/online_help/index.html?hid_web_maxdiff_badrespondents.htm), was there any numbers for 99% correct classification available?

I've got a max-diff survey of ocean scientists with 2187 completes (4 items per question, 36 comparisons in total to rank a list of 67 research priorities)- my lowest RHL score is 305, so responses are fine in aggregate with nobody appearing really random. If possible, I'd just like to be able to state that the 99% cutoff would be and see how many respondents fall in the 95-99% gap.

asked Mar 23, 2014 by murray rudd

1 Answer

0 votes
Hi, Murray.

You can replicate this analysis to fit the specifics of your experiment.  Using the "generate data" function in SSI Web version 8 just run 1,000 random responders through the MaxDiff section of your survey.  Then run HB analysis on your random responders and sort the resulting HB utility file in order of RLH.  You can now investigate any properties of the RLH distribution that interest you - mean, median, n-tiles, whatever.
answered Mar 23, 2014 by Keith Chrzan Platinum Sawtooth Software, Inc. (50,675 points)
Great - thanks very much.