Dave and I don't always agree :-) I don't think we've got either the database of "people not attending because of costs" nor a good model for factoring them in if we did (e.g. N pnac's times some percentage who would still not attend because of other issues times some percentage where the location is problematic * etc etc). If you can figure this out, I think you could probably apply the same model to forecasting the US stock market... We have good data on past attendees. From that we can probably build a pretty good model on what each past attendees Pa (percentage chance of attending the next meeting) is. From that we can probably build a pretty good model of what our probable attendee demographics will look like absent the one-shot and/or local attendees. Let's stick with solid data rather than try and resolve the hypotheticals - I doubt the latter is possible in any meaningful way. Mike At 08:52 AM 9/6/2010, Dave CROCKER wrote: >On 8/30/2010 1:10 PM, Melinda Shore wrote: >> There's already bias in the population >> of meeting attendees - do you, and, if so, how do you account >> for people who are already not attending because of costs? > >Yes! This is exactly the point I keep raising, about sampling error. We need to make sure that our analyses are with respect to the population (the full set of potential attendees) that we have in mind, rather than just a core of folk who usually attend. > >A theoretical analysis of fair costs doesn't query real people, so sampling is not an issue. But the second we start surveying, we need to consider folks not on the ietf@ or ietf-attendees@ lists. _______________________________________________ Ietf mailing list Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf