Wednesday: Here's largely how my day went.

I guess I should at least be happy that I've checked two things off the list for this week.

Here's a bear picture, because that will hopefully push that gif off the compose screen, since it's kind of giving me a headache.

Bear doesn't care.  He's got a whole river to explore.

I also discovered a stats bug I should have foreseen.  In the program, I calculate the running median and the running MAD-based sigma proxy.  To save time, I do this as a single pass through the data, and then do some logical scaling of sigma after the fact based on whether we plan to subtract or divide the median values from the data.  However, this is only acceptable when the best fit slope to the median is effectively zero.  When that's not the case, the measured sigma is actually \sigma_M^2 = \sigma_true^2 + \sigma_slope^2.  \sigma_slope is something like (m * w * 0.37), where m is the linear slope, w is the width over which the running stats are done, and the number is based on the Gaussian CDF and blah blah other stuff.  Since most of the data I've been working with have m on the order of 1e-3 and w on the order of 1e1 or so, this hasn't been much of a problem.  However, today I was looking at things with m ~ 10, so all noises were scaled by something like a factor of four or so.

I really think Charles Krauthammer is a dick, but this is just uncalled for, ask.com.

• Don't click this link, it's super depressing.  If you do click this link, definitely do not scroll down to the comments to see the other Animal Crossing comic.  It's even worse.
• Wow, after that last one, Nurse Joy being mean isn't that bad.  I think between the two links, I maybe don't want to get a 3ds so I can play games.
• I thought that the original one was really far away, but it turns out that it's about the same distance from my house as work is.  I should stop by over the weekend, I think.