This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: STA 2006 Hypotheses Testing (Section 8.2 and 8.3) 20072008 Term II 1 Simple Hypothesis vs Simple Hypothesis Suppose X 1 ,...,X n are sampled i.i.d. N ( μ,σ 2 ) with known σ 2 . • We know μ can only be either μ or μ 1 . ( μ 1 > μ ) . That is, H : μ = μ vs. H 1 : μ = μ 1 • H is called the null hypothesis . • H 1 is called the alternative hypothesis . • Task: Determine which hypothesis is more ”true” based on the sample information. A legitimate decision rule is to conclude H 1 is more true (reject H ) if ¯ x n ≥ c where c is a real number. We are bounded to commit two types of error: H is true H 1 is true Reject H Type I error No Error Reject H 1 No Error Type II error 1 2 1. Simple Hypothesis vs Simple Hypothe sis The corresponding probabilities are: Let Z ∼ N (0 , 1). α = Pr( Committing Type I error ) = Pr μ = μ ( ¯ X n ≥ c ) = Pr( Z ≥ c μ σ/ √ n ) = doubly crossed area in the figure (*) β = Pr( Committing Type II error ) = Pr μ = μ 1 ( ¯ X n < c ) = Pr( Z < c μ 1 σ/ √ n ) = singly crossed area in the figure. • We can decrease α by increasing c but then β will be increased. • Similarly, β can be decreased by choosing a smaller c but then α will be increased. • Conclusion: There is no way protecting both hypotheses with equal emphases. • Statisticians decide to protect H (”presumed innocence”) such that α is kept to be a small number (usually 0.05) and c can be fixed accord ingly. 2 • How to fix c ? By ( * ) , Pr( Z ≥ c μ σ/ √ n ) = α ⇒ c μ σ/ √ n = z α i.e. c = μ + z α σ/ √ n . Therefore, the decision rule becomes: if ¯ X n ≥ μ + z α σ/ √ n, reject H . • Interpretation: If such decision procedure is repeated many times, only 5% of the time H will be wrongly rejected. • A more formal way of writing the decision rule is: Reject H if ¯ x n μ σ/ √ n ≥ z α ....
View
Full
Document
This note was uploaded on 06/05/2011 for the course STATISTICS 2006 taught by Professor Ho during the Spring '11 term at CUHK.
 Spring '11
 Ho
 Statistics

Click to edit the document details