In article <[EMAIL PROTECTED]>, zhutou <[EMAIL PROTECTED]> wrote: >> In article <[EMAIL PROTECTED]>, >> Glen <[EMAIL PROTECTED]> wrote:
>> You cannot CHECK this, but you can TEST it. One problem is that >> there are an infinite number of tests, and you can be sure that >> some of them will fail, even if the sample is really iid. >> There are many situations in which this is done. >So >(1) how can we always assume that? For example, I get a set of sample >data by obeservation, then I assume i.i.d, then I can draw very far >conclusions from these data involving some complex calculations; >(2) could you give me some pointers to the books discussing this >topic? Consider the simplest situation, in which all the observations are 0 or 1. If they are iid, then the probability that each observation is 1 is some number p between 0 and 1. If this is all that is available, this cannot be tested by classical methods. All that can be tested in that manner is whether the location of the 1's is random. One test is whether the number of 1's in the first half is "random". Another test would look at the transitions. Bayesian methods can, at least in principle, do everything. -- This address is for information only. I do not claim that these views are those of the Statistics Department or of Purdue University. Herman Rubin, Deptartment of Statistics, Purdue University [EMAIL PROTECTED] Phone: (765)494-6054 FAX: (765)494-0558 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
