Statistics Week 4

This week's homework lived up to it's description of "really tough".  That said, I had a good time doing it.  I actually got to break out the pen and paper again and do some algebra.  And darn it if algebra isn't just fun.  I'm not completely sure I got the right answers, and I'll be interested to see what my group members came up with, but I had a good time crunching the numbers.  It started to feel like I could really start applying this to some of the numbers we get at school and bring some more meaning to them.

I was also impressed with the Bill Tancer video.  There was one question from the audience that I didn't discuss on the wiki, mostly because I wanted to talk about it here.  A member asked how he determined the background of the participants in his data collection- how did he know which people searching for Stacy Kiebler were male, and which were female.  He answered with a long string of terms that were Greek to me, but it seemed like he gave people the chance to voluntarily submit their information.  The audience member asked, basically, doesn't that skew the data?  Aren't you just analyzing the subset of the population that's actually willing to hand over their information (which he said 90% of the people at Google wouldn't do)?

It made me think of the Nielsen ratings.  A few years ago, I actually filled out the little Nielsen packet, mostly because I collected a teacher's paycheck and $20 sounded awesome.  I caught myself skewing my own data, though (something I alluded to in the wiki).  I caught myself watching stupid shows because I was bored and it was all that was on, but I didn't actually want to encourage the production of those shows, so I claimed that I watched nothing.  Other times, I would see something that was about to come on after my bedtime that looked really interesting, so I claimed that I watched it to encourage more of those shows to be made.  (Yup, I just admitted to liking algebra and having a set bedtime in the same blog post.  Awesome.)  I wondered how many other people did that- gave Nielsen slightly inaccurate data to encourage a particular result in the survey.  I'm sure statisticians have found a way to account for this, but my gut tells me it's probably not quite perfect.  Perhaps these small changes just get accounted for in a large enough sample size?  Either way, it's something that will be in the back of my mind as we go through our own survey data.

Oh, and if anyone from Nielsen is reading and wants their $20 back, I was just kidding.  I totally watched that NatGeo special.