I've been playing with a dataset that gives slightly, but importantly different results across a range of values for dc. This prompted me to walk through the estimateDc code to make sure I understood how it was generating a value. It seems that it is iteratively finding a distance that lies between neighborRateLow and neighborRateHigh (default 1-2%) of the distances.
I've had a few cases where estimateDc enters an endless loop. Do I understand it correctly that I am producing a similar estimate if I just use quantile(dist, 0.015), which does not hang? On the handful of datasets that I've tried, this produces a value very close to what I get with estimateDc, but I want to make sure that I understand properly that estimateDc is just a faster way of getting an approximate quantile.
I've been playing with a dataset that gives slightly, but importantly different results across a range of values for
dc. This prompted me to walk through theestimateDccode to make sure I understood how it was generating a value. It seems that it is iteratively finding a distance that lies betweenneighborRateLowandneighborRateHigh(default 1-2%) of the distances.I've had a few cases where
estimateDcenters an endless loop. Do I understand it correctly that I am producing a similar estimate if I just usequantile(dist, 0.015), which does not hang? On the handful of datasets that I've tried, this produces a value very close to what I get withestimateDc, but I want to make sure that I understand properly thatestimateDcis just a faster way of getting an approximate quantile.