Here is what I accomplished (about 2 weeks ago)… six dichotomous choice contingent valuation logit curves with the probably of a yes (willing to pay the bid) on the vertical axis and the bid amount on the horizontal axis. Three of the data sets have a “fat/flat tails” problem and three don’t.
Head’s up, you can’t do much in the way of hypothesis testing if your data suffers from fat/flat tails because your standard errors are too wide (and asymmetric). If you estimate WTP with the nonparametric Turnbull you’ll get tighter and symmetric confidence intervals but you’re cheating, especially if your data is smoothed to hide those non-monotonicities. More on this in about 2 years …
