This is a replication of a question I've recently asked on Cross Validated. It hasn't received an answer or much attention, so I've posted it here.
I have a family of point processes representing neural firing data. In each of these point processes, there is a marked pause in events beginning and ending at around the same time. I would like to measure the length of this pause.
The model generating this data seems quite complicated and, so far, it has resisted fitting to any elementary distributions. This makes changepoint detection difficult. I've developed a non-statistical method inspired by Canny edge detection, though it doesn't work as well as I'd like. Are there known methods for detecting changepoints in general, nonstationary point processes?