I can't sleep...

It's 3am in the morning and I can't sleep. I'm sitting here in bed with my laptop on my comfortor thinking, "Why?" There's no reason for this. I'm absolutely exhausted and yet I can't sleep?!? Perhaps I'm stressed about tomorrow or that fact that I have to get up early. I have to meet my prayer partner at 9am now because I have a class that starts at 11am. We used to meet at 10am. I've never had to get into school that early. And I'm not a morning person at all. If I have to get up very early, my mind and body feels like there's no reason for me to even go to bed. Plus, I think I'm getting all worked up over the stuff I have to do tomorrow. I'm holding a research meeting in which I'm the lead. I've done this many times before, and I spent 3 or 4 hours this evening working on it in preparation. But, I feel like I'm winging it. In one sense, I'm really excited about it. In another sense, I'm stressed about it going well. I did't get a chance to read all the articles I'm handing out to the others in the group. So I'm not positive how helpful they will be. I talked with another grad student earlier today and invited him to join our project as well. So now there is four of us total. I even wrote up a little page and a half, mini proposal for the project. I just feel like I'm missing something or maybe I'm concerned that the undergrad is going to make fun of me (in a teasing way, he's just like that) or act up and mimic our advisor instead of considering this a serious work effort. I don't know if he's going to help with the data collection or not. I haven't printed out articles that are specifically related to his previous work that we are planning to write up into an article.

The other thing I'm worried about is my methods class on IRT. We're learning about Guttman scales and my prof spent the whole previous class period trying to present crazy examples and get us to derive our own conclusion as to what is wrong with the example. That made me a little uncomfortable. All of my previous stats classes have been very detailed and axplanatory. We've worked through problems in class and I've been able to follow it well. But, "discovery learning" in a stats class? We're not statistically minded enough to do that!! So my prof told us to play with the numbers and come back to class today with our own method for counting errors. Uggh! Then he only suggested we read the article again. Well, I read it again and now In understand about 10% more of it than I did last time, which was only about 15% of it anyways.

So since I can't sleep, let me try to explain Guttman scaling. Maybe it will help me to explain it. No, it's not working. Here, instead, let me try listing the the words I don't understand. Minimal margainal error. What is that and why is it like PRE (proportional reduction in error?). What is PRE anyways. Lets try coefficient of reproducability. Why can't I remember what that means? Why do they have to use the term reproducabilty? Okay, I pulled out the article. It says, "the amount by which a scale deviates from the ideal scale pattern is measured by a coefficient of reproducability. This coefficient is simply a measure of the relative degree with which the obtained multivariate distribution corresponds to the expected multivariate distribution of a perfect scale." Okay, so it's a measure of how well my data that I collected corresponds or reproduces the ideal data. Ideal data has line up with the cumulative model, which means that the easiest questions come first and so on. But, there are many response patterns that aren't going to fit the ideal. We don't want these, noooo. This is error. So our coefficient of reproducability measures the extent to which our data reproduces an ideal scale, which is unlikely to happen in the real world due to random error or potentially a second underlying scale. Okay, so what does that have to do with minimal marginal reproducability? So apparently the coefficient of reproducability (CR) is not an accurate measure of the extent to which our data reproduces the ideal. This is because the CR can be, wait, backing up. There is a cut off in which we accpet our data if the amount of error is less than 10% of the possible error. But, apparently if we have extreme values, umm, marginals, I think, then the CR isn't accurate. This happens because somehow the extreme values make it so that our CR is very small, but there values make it so that it can only be a certain extent. If it is only possible to have 5% error and you get 4% error, that means you have a lot more error than if you have 60% chance of error and you you only get 9% error. So now you have to compare the CR to the MMR, which is somehow a measure of the possible total error, I think. But, then the article uses the words marginal error and nonmodal category, and I don't understand what those are, or marginal frequency. See I want him to work out an example on the board, pointing to the numbers and equations that stand for the words that the article drops without explaining. But, I just know he's not going to do that, because he doesn't operate that way.

Well, now it's 3:36 and my brain really just isn't working at all. I might trygoing to sleep now and maybe I'll be successful. No, everything is still churning around in my brain. Brain, why don't you turn off!