http://www.talkstats.com/ Statistical theory and methodology. Mathematical statistics. Parametric inference. Nonparametric inference. en Tue, 18 Jun 2013 20:04:51 GMT vBulletin 60 http://www.talkstats.com/images/misc/rss.png http://www.talkstats.com/ http://www.talkstats.com/showthread.php/44568-Am-I-applying-this-Chi-squared-test-correctly?goto=newpost Sun, 16 Jun 2013 01:26:12 GMT Hi everyone! I am a history grad student and I am trying to use some statistics in one of my papers. I am working with a certain (ancient) text... Hi everyone!

I am a history grad student and I am trying to use some statistics in one of my papers. I am working with a certain (ancient) text and am trying to argue that its last quarter is significantly different from its first three quarters fifths in several textual measures (the null hypothesis being that the whole text is supposed to be similar). I'll provide a real example:

This text is 498 pages long (378 pages in part A and 120 pages in part B). Part A has overall 10 uses of first person throughout it, while part B has 27. I thought that I could apply the chi-square test to see if these parts are indeed independent from each other (i.e are written differently), and created this 2x2 matrix:

378 | 10
------------
120 | 27

The result of this test is statistically significant, but I am very uncertain if I used it correctly. I'll be glad to answer any other questions.

Thank you so much! ]]> Statistical Research badslacker http://www.talkstats.com/showthread.php/44568-Am-I-applying-this-Chi-squared-test-correctly http://www.talkstats.com/showthread.php/44567-test-of-equality-of-covariance-(or-correlations)-among-DISCRETE-variables?goto=newpost Sun, 16 Jun 2013 01:25:37 GMT i have been trying to find significance tests whose null hypothesis is that the non-redundant, off-diagonal elements of a covariance (or correlation)... i have been trying to find significance tests whose null hypothesis is that the non-redundant, off-diagonal elements of a covariance (or correlation) matrix are all equal. i've arlready tried 3 tests of the covariance matrix and one for the correlation matrix. they all work wonderfully for continuous data, but the moment i try them with discrete data, they brake down in terms of type 1 error rate of .05 (most turn out to be very conservative).

i'm really at my wits end here. has anyone heard of a test with the null hypothesis that i mentioned previoulsy but which preserves its type 1 error rate (of .05) for DISCRETE data? ]]> Statistical Research spunky http://www.talkstats.com/showthread.php/44567-test-of-equality-of-covariance-(or-correlations)-among-DISCRETE-variables http://www.talkstats.com/showthread.php/44565-True-mean-and-the-Central-Limit-Theorem?goto=newpost Sat, 15 Jun 2013 20:56:06 GMT I've learned some pretty remarkable things about the Central Limit Theorem and the so-called Sampling Distribution of the Sample Mean. But there's one thing I've heard that I am pretty skeptical about. Can I treat the Sampling Distribution of the Sample Mean as a probability density function of the true mean?



As the number of samples and for some finite sample size , does the above actually hold?

I know that for very small , the Sampling Distribution of the Sample Mean may not actually look like a normal distribution at all. So I guess my question applies only for large enough that skew and kurtosis are small. ]]> Statistical Research aut0pilot101 http://www.talkstats.com/showthread.php/44565-True-mean-and-the-Central-Limit-Theorem