To: nnanos@nanosresearch.com, fgraves@ekos.com, lbozinoff@forumresearch.com
hi.
i don't think it's going to require much of an argument from me to have you acknowledge that provincial wide data isn't very useful to plug into riding modelling the way it's being done; the idea that a swing in vancouver might be felt in kelowna, or that a swing in montreal might be felt in sherbrooke, is not really well thought out. canadian provinces are big and diverse and often have two or three proper sample frames, as opposed to american states, which are mostly small and get away with one, or maybe two if you're being really rigorous. if people doing models are to create systems that are robust enough to handle the massive and unpredictable swings we're seeing in this election, they're going to need more targeted sample frames to begin with.
but, you don't model; you poll. maybe, if a few of you might want to think about modelling, we could get some better sample frames.
for right now, i know you're focused on what you're doing. but, releasing data organized by area code would probably be a five minute job, right? organize by area code. print. done...
in lieu of better sample frames for the models, maybe it might help people understand things a little bit better, so they're not being led astray by models that are operating well beyond the assumptions they were constructed with.
j