testing multinomials

Hi, everyone,

I have a question about testing multinomial distributions, and would like
hear your comments.

Let X = (x_1, x_2, ..., x_k) be an observation from a multinomial
distribution f_1(p_1, p_2, ..., p_k), and another observation Y = (y_1, y_2,
...., y_k) from another multinomial distribution f_2(q_1, q_2, ..., q_k),
where p_i and q_i are UNKNOWN, now can we design a test statistic to test
H_0: p_i = q_i, vs. H_1: p_i != q_i?

In other words, we want to test whether X and Y came from two multinomials
that are roughly the same. In case it is needed, we can take k = 4.

thanks and any comments or pointers are appreciated!
Chao



.

Relevant Pages

  • testing multinomials
    ... I have a question about testing multinomial distributions, ... where p_i and q_i are UNKNOWN, now can we design a test statistic to test ... Chao ...
    (sci.stat.math)
  • Re: testing multinomials
    ... I have a question about testing multinomial distributions, ... q_k), where p_i and q_i are UNKNOWN, now can we design a test statistic to ... Chao ...
    (sci.stat.math)
  • Re: testing multinomials
    ... chao liu wrote: ... I have a question about testing multinomial distributions, ... where p_i and q_i are UNKNOWN, now can we design a test statistic to test ...
    (sci.stat.math)