11 October 2008

Statistically Significant

I finally got my statistical program to work. Ha! Power...

Anyway, I suppose the previous paragraph did not make sense. I am doing a multidimensional scaling analysis on a bunch of data about a couple hundred of languages for a seminar paper. Until this week, I couldn't make the statistical program to run, but somehow, I found the solution to my problem.

Apparently, I have too many samples. I am sampling languages that have a specific characteristic, and seeing how the topology of those languages are; whether some languages are more similar to one language than another. For that, I am using ALSCAL techniques to measure the similarity difference between my samples. Since my data are binary-coded, I am using a binary classical MDS test to measure the Euclidian distance between all of my samples, and therefore create a two-dimensional topological space of the similarity of my data.

And from that, interesting results are coming out. Therefore today, I started to sit down and write this seminar paper. Good thing, so that I could take that out of my plate. And now I could say that I know another statistical trick under my belt.

I suppose that is a good thing. I now know the regular vanilla descriptive statistics, the regular parametric inferential methods, such as chi-squares, t-tests, F-tests, and so on. I also know how to construct the different flavors of ANOVAs, and I also have experience running binary logistic regressions. Now I know how to do multidimensional scaling on binary-coded data.

Perhaps the most important thing when one does these statistical tests is that one fully understands what the numbers mean. The thing is, with a given program, one can just feed the program some data, and then click on a few buttons with a mouse, and voila! The program spews out a bunch of numbers and there you go!

Anyway, I am looking forward to tomorrow. A friend and I are going to visit this mansion somewhere on the outskirts of Rochester. There is this mansion called the Sonnenberg Mansion, and it features a greenhouse and plenty of gardens. It apparently was featured in a documentary called America's Castles. So I am bringing my camera and enjoying this brief respite from the academic work I am currently dirtying my hands in.

(Dead Trees Among the Dead, from my Arlington Cemetery Series)


  1. I absolutely hate stats so I wish you good luck. I tried to understand how it relates to linguistic and I guess it does make sense... linguistic is somehow very mathematical.

  2. Zhu,

    Well, let me give you a simple example. Say we want to know how English speaker use the verb "perform", could it only be used with a count noun as an object, or even mass or abstract nouns? One could do a corpus survey and look at samples of written and spoken data, and could see the proportion of both uses. So one can then see that "perform" is predominantly a verb that takes a count noun object, like "perform a trick" but one could also do "perform magic".