thoughtcrumbs

Updates

I've been playing catch up ever since returning from Iceland, so this blog has gone horribly neglected. But my mushroom stock is simmering for forty minutes now, giving me a brief moment for navel-gazey updates. If you're not my mother, you can probably skip this post.

• Iceland first. I went there for a week with friends from the lab, same group that went to Ecuador last year. We're continuing our goal to hit a different continent every year until graduation. (Iceland counts as Europe, although the continental rift with North America passes right through the island. So we could always go to, say, Prague later.
  
  Highlights of the trip included cross-country skiing, daily bathing in thermal springs, kayaking through a frozen lake (requiring cracking the ice with your paddle, in a sortof full-body creme brulee motion) and building and sleeping in an igloo. Our guides, the magnanimous Runar, Siggy, and Marij of Borea Adventures were somewhat surprised, I think, at our decided intrepidness. Iceland pictures are up at Flickr.
• I put my Hawaii trip photos up on Flickr, too -- completely forgot about them. Josh and I spent the first week of January there, and I definitely attended every minute of the conference. Definitely.
• I'm writing to another blog that will launch in about a month. More deets soon.
• It's still cold in Pittsburgh, after a flirty week in the 60s, in which all the crocuses and tulips bloomed. But it's expected to freeze again this weekend, meaning it's still too early to start my porch herb garden. The basil's probably knee-high in Oregon by now, but everything's delayed in Pennsylvania. I resorted to splurging on one of the $5 thyme pots at Whole Foods and don't feel particularly sheepish about it.
• On the research front, I've discovered a fantastic new statistical model: zero-inflated negative binomial. It's perfect for data like mine (# of replies to newsgroup posts), where you have both overdispersion and a whole lot of zeros. But it's a pain to run, confusing to analyze interactions, and will be challenging to explain in publications. But it's worth it to have an appropriate model (and thus good results). I'm thinking about starting another project, automatically detecting politeness in online discussion groups. There's a well-known twenty-year-old model of politeness markers that I'd like to apply to current online discussions, and hopefully train a machine learning algorithm to pick out those markers. Then I'll check to see if politeness matters (e.g. in political groups, maybe you get more replies if you're less polite).
• Going to New Orleans again in May to interview musicians about changes to their careers after Katrina. We interviewed about 50 people last year, and are scheduling follow-ups now. Just watched the Spike Lee documentary, When the Levees Broke, which was just heartbreaking. New Orleans is so full of vigor and talent, and I'm hoping the city's renovation reflects that a little more this year than last. But I'm not expecting the lower ninth to look much different, sadly.