Thanks for starting the JMP forum! Looks like I'm the first non-SAS submitter, so...
I'm trying to develop factors to add to an existing likelihood-to-buy model. I'm taking known sales and create mean and stddev on time between purchases, plus a variable representing time-since-most-recent-buy. Is there a platform in JMP that could take the purchase info (one record per item purchased) and generate these stats on a per-customer basis?
I've searched JMP help for periodicity, and the closest I've seen so far is Survival and Reliability, Recurrence Analysis, but I can't see how to adapt the required Cost, End of Service field for my purposes. Of course, I can use SAS and SQL to construct the data I want, but if there's a tool that'll bypass that tedium, I'd love to use it.
Given what you say above, Steve, the chances are a response would come from a SAS person, and just to be clear, I am one such – Actually I work with JMP specifically. If you want or need the convenience of working in one environment, then it’s perfectly possible (through JMP’s scripting language, JSL) to do this (at least for modest data volumes). Not sure if it’s exactly what you need, but I recently submitted an entry to the File Exchange at http://www.jmp.com/community/ that may help (there can be bit of a lag because of the review process).
To move the discussion on a little, my motivation for the above was to find some useful ways to visualise raw point of sale transactional data prior to more formal statistical modelling. Because we are all consumers in one way or another, it’s easy to appreciate that the behaviour that drives the dynamics of the buying process is non-trivial. At any rate, it seemed to me that by using summary data prematurely we are open to what’s been called ‘the flaw of averages’, which is what led me to investigate fruitful ways to look at the raw data. One idea (featured in the script) is to use a parallel coordinates plot,