This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Lecture 24 24.1 Goodnessoffit test. Suppose that we observe an i.i.d. sample X 1 , . . . , X n of random variables that can take a finite number of values B 1 , . . . , B r with some unknown to us probabilities p 1 , . . . , p r . Suppose that we have a theory (or a guess) that these probabilities are equal to some particular p ◦ 1 , . . . , p ◦ r and we want to test it. This means that we want to test the hypotheses H 1 : p i = p ◦ i for all i = 1 , . . . , r, H 2 : otherwise, i.e. for some i, p i 6 = p ◦ i . If the first hypothesis is true than the main result from previous lecture tells us that we have the following convergence in distribution: T = r X i =1 ( ν i np ◦ i ) 2 np ◦ i → χ 2 r 1 where ν i = # { X j : X j = B i } . On the other hand, if H 2 holds then for some index i , p i 6 = p ◦ i and the statistics T will behave very differently. If p i is the true probability ( X 1 = B i ) then by CLT (see previous lecture) ν i np i √ np i → N (0 , 1 p i ) . If we write ν i np ◦ i √ np ◦ i = ν i np i + n ( p i p ◦ i ) √ np ◦ i = ν i np i √ np i + √ n p i p ◦ i √ p ◦ i then the first term converges to N (0 , 1 p i ) but the second term converges to plus or minus ∞ since p i 6 = p ◦ i . Therefore, ( ν i np ◦ i ) 2 np ◦ i → + ∞ 94 LECTURE 24.LECTURE 24....
View
Full
Document
This note was uploaded on 10/11/2009 for the course STATISTICS 18.443 taught by Professor Dmitrypanchenko during the Spring '09 term at MIT.
 Spring '09
 DmitryPanchenko
 Statistics

Click to edit the document details