When dealing with irregularly spaced quote data there are often ghost quotes.. people are fishing for hidden orders and they last less than a few ms. This can create noise it not handled properly in different situations: For example, when creating a reguarly space time series it is common to find the average price in that interval, or in the case of returns the sum of returns during an interval. The problem with this is if there are relatively few "real" quotes with a long duration, and many more fake quotes with a very short duration then the average quote will be skewed by the existince of all the fakes since they are simply averaged. I believe some people might be acting on this and that is why fishing sometimes works.. A solution I thought of is to weight the quotes according to the length of the duration of the quote, 1-2 ms durations would be almost entirely eliminated and thus would reduce the skew... It might be possible to somehow weight in the size of the quote too, but it is not apparant how useful that would be. Thoughts, is this useful, anyone doing something similar?
It's similar concept to time weighted average spread which tries to get a idea of the mean spread without the influence of short-lived quotes.
I'm not, this is for very liquid stocks. I'm just a perfectionist and don't want to introduce any bias into my system.