My Canadian compadres are really starting to drive me nuts. I owed them a laundry list of items to support the validity of the mathematical model we developed for Canadian Indoor reach and frequency. Peeled off a couple of them myself, then pounced on Gary to throw together a few graphs to round off the list. Found a mistake in some of the graphs, in that they referenced not GRPs as the x-axis, but something else that made the curves for the empirical data vs. our regressions look farther off than they really are, but he fixed them, and they look as though the pink and blue dots are practically on top of each other, as I knew they should be, from our r-squared values.

If this is zooming over your head faster than a jet, let me step back and clarify that I head up the Advanced Analytics department within my company. We make a living doing math and statistical modeling. The funny thing is that conventional wisdom holds that girls are lousy at math and science. Wrong. Two-thirds of my department are women.

Our client's complaints are numerous:

• "Your regression statistics are too good to be true."
• "Why didn't you alert us to all the holes in the data we commissioned from TNS/Canadian Facts?"
• "We were fine with your use of a beta binomial frequency distribution last week, and the week before, but we now demand that you use gamma poisson instead."

Yeah, the r-squareds are phenomenal, but it's not unusual for them to be good when you have smooth data from which to run your regressions. It's not our job to spend weeks looking for every hole in the data you commissioned--that's the data supplier's job. We can list them, and it seems like a lot, but yes, there will be holes when you look at demographics such as older age breaks for the college/university campus indoor advertising network, as well as no data at all for non-students. You'll also find holes in upper income breaks for college students. These are not data anomalies--they'd be anomalies if the data existed! Well, if you insist on us implementing a gamma posson frequency distribution, it'll cost you more in terms of data analysis on our part, because this was never covered in the contract, and you've already gotten an extra 2.5 weeks worth of freebie data analysis from us that was never covered in the contract, either.

People--get a grip. We're good at what we do. If there were data that made no sense, that would be a problem, but where it doesn't exist makes perfectly good sense to us.