Вы находитесь на странице: 1из 3

Hypotheses

Chapter 8 8.1 The null hypothesis, denoted H0, is the A Test of Hypotheses
Tests of Hypotheses and
claim that is initially assumed to be true.
The alternative hypothesis, denoted by A test of hypotheses is a method for
Hypotheses Based Test Procedures Ha, is the assertion that is contrary to H0. using sample data to decide whether
the null hypothesis should be
Possible conclusions from hypothesis-
on a Single Sample testing analysis are reject H0 or fail to
rejected.
reject H0.

Test Procedure
Errors in Hypothesis Testing Rejection Region: α and β
A test procedure is specified by Significance Level
1. A test statistic, a function of the
A type I error consists of rejecting the Suppose an experiment and a sample Specify the largest value of α that
sample data on which the decision is
null hypothesis H0 when it was true. size are fixed, and a test statistic is can be tolerated and find a rejection
to be based.
chosen. The decreasing the size of the region having that value of α. This
2. A rejection region, the set of all test A type II error involves not rejecting H0 rejection region to obtain a smaller makes β as small as possible subject
statistic values for which H0 will be when H0 is false. value of α results in a larger value of β to the bound on α . The resulting
rejected (null hypothesis rejected iff for any particular parameter value value of α is referred to as the
the test statistic value falls in this consistent with Ha. significance level.
region.)

Case I: A Normal Population Case I: A Normal Population


Level α Test 8.2 With Known σ With Known σ
A test corresponding to the significance Alternative
level is called a level α test. A test with Tests About a Null hypothesis: H 0 : µ = µ0 Hypothesis
Rejection Region
for Level α Test
significance level α is one for which the Population Mean H a : µ > µ0 z ≥ zα
type I error probability is controlled at x − µ0
the specified level. Test statistic value: z= H a : µ < µ0 z ≤ − zα
σ/ n
H a : µ ≠ µ0 z ≥ zα / 2 or z ≤ − zα / 2

Recommended Steps in Hypothesis-Testing Analysis Hypothesis-Testing Analysis Type II Probability β ( µ ′) for a Level α
Hypothesis-Testing Analysis 4. Give the formula for the computed
Test Type II
7. Decide whether H0 should be Alt. Hypothesis Probability β ( µ ′)
value of the test statistic.
1. Identify the parameter of interest and rejected and state this conclusion in µ0 − µ ′
describe it in the context of the H a : µ > µ0 Φ zα +
5.State the rejection region for the the problem context. σ/ n
problem situation. selected significance level
H a : µ < µ0 µ0 − µ ′
The formulation of hypotheses (steps 2 1 − Φ − zα +
2. Determine the null value and state the σ/ n
null hypothesis. 6. Compute any necessary sample and 3) should be done before examining
quantities, substitute into the formula the data. H a : µ ≠ µ0 Φ zα / 2 +
µ0 − µ ′ µ −µ ′
− Φ − zα / 2 + 0
3. State the alternative hypothesis. for the test statistic value, and compute
σ/ n σ/ n

that value.
Sample Size Large Sample Tests (n > 40) Case III: A Normal Population
Case II: Large-Sample Tests Distribution
The sample size n for which a level α test
also has β ( µ ′) = β at the alternative value
For large n, s is close to σ.
When the sample size is large, the z If X1,…,Xn is a random sample from a
µ ′ is X − µ0 normal distribution, the standardized
tests for case I are modified to yield Test Statistic: Z =
σ ( zα + z β )
2
valid test procedures without requiring S/ n variable
one-tailed test X −µ
either a normal population distribution T=
µ0 − µ ′ The use of rejection regions for case I S/ n
n= or a known σ .
2 results in a test procedure for which the
σ ( zα / 2 + z β ) significance level is approximately α .
has a t distribution with n – 1
two-tailed test degrees of freedom.
µ0 − µ ′

The One-Sample t Test The One-Sample t Test A Typical β Curve for the t Test
8.3
Null hypothesis: H 0 : µ = µ0 Alternative β curve for n – 1 df
Hypothesis
Rejection Region
for Level α Test β when Tests Concerning
x − µ0 µ = µ′ a
Test statistic value: t= H a : µ > µ0 t ≥ tα ,n −1
s/ n 0
H a : µ < µ0 t ≤ −tα ,n −1 Value of d corresponding to
Population Proportion
H a : µ ≠ µ0 specified alternative to µ ′
t ≥ tα / 2,n −1 or t ≤ −tα / 2,n −1

Large-Samples Concerning p
Large-Samples Concerning p
A Population Proportion Large-Sample Tests Alternative Rejection Region
Hypothesis
Let p denote the proportion of Null hypothesis: H 0 : p = p0 H a : p > p0 z ≥ zα
Large-sample tests concerning p are
individuals or objects in a a special case of the more general
population who possess a specified Test statistic value: H a : p < p0 z ≤ − zα
large-sample procedures for a
property. parameter θ . pˆ − p0 H a : p ≠ p0
z= z ≥ zα / 2 or z ≤ − zα / 2
p0 (1 − p0 ) / n Valid provided
np0 ≥ 10 and n(1 − p0 ) ≥ 10.

General Expressions for β ( p′) General Expressions for β ( p′)


Sample Size Small-Sample Tests
Alt. Hypothesis β ( p′) The sample size n for which a level α test Test procedures when the sample size
H a : p > p0 Alt. Hypothesis β ( p′) also has β ( p′) = p
p0 − p ′ + zα p 0 (1 − p 0 ) / n n is small are based directly on the
Φ
p′(1 − p′) / n p0 − p ′ + zα p 0 (1 − p 0 ) / n zα p0 (1 − p0 ) + z β p ′(1 − p ′)
2
one-tailed binomial distribution rather than the
H a : p ≠ p0 Φ test normal approximation.
p′(1 − p′) / n p′ − p0
H a : p < p0 p0 − p ′ − zα p0 (1 − p0 ) / n n=
1− Φ
p′(1 − p′) / n zα / 2 p0 (1 − p0 ) + z β p ′(1 − p ′)
2
two-tailed P(type I) = 1 − B(c − 1; n, p0 )
p0 − p ′ − zα p0 (1 − p0 ) / n
−Φ p′ − p0 test
p′(1 − p′) / n B( p′) = B(c − 1; n, p′)
P - Value P - Value P-Values for a z Test
The P-value is the smallest level of
The P-value is the probability, calculated
significance at which H0 would be P-value:
8.4 assuming H0 is true, of obtaining a test
rejected when a specified test procedure
statistic value at least as contradictory to
is used on a given data set. 1 − Φ(z) upper-tailed test
P - Values 1. P -value ≤ α
H0 as the value that actually resulted.
The smaller the P-value, the more P= Φ( z) lower-tailed test
reject H 0 at a level of α contradictory is the data to H0.
2 1 − Φ( z ) two-tailed test
2. P-value > α
do not reject H 0 at a level of α

P-Value (area) Constructing a Test Procedure


P-value = 1 − Φ ( z )
Upper-Tailed P–Values for t Tests 8.5 1. Specify a test statistic.
P-value = Φ ( z ) 2. Decide on the general form of the
0 z
The P-value for a t test will be a t Some Comments on rejection region.
Lower-Tailed curve area. The number of df for the
one-sample t test is n – 1. Selecting a 3. Select the specific numerical critical
-z 0
P-value = 2[1 − Φ (| z |)] Two-Tailed
Test Procedure value or values that will separate the
rejection region from the acceptance
region.
-z 0 z

Issues to be Considered Issues to be Considered Statistical Versus Practical The Likelihood Ratio Principle
1. When there exist two or more tests that Significance 1. Find the largest value of the likelihood
1. What are the practical implications and
are appropriate in a given situation, how for any θ in Ω0 .
consequences of choosing a particular Be careful in interpreting evidence when
can the tests be compared to decide
level of significance once the other the sample size is large, since any small 2. Find the largest value of the likelihood
which should be used?
aspects of a test procedure have been departure from H0 will almost surely be for any θ in Ωa .
determined? 2. If a test is derived under specific detected by a test (statistical significance),
assumptions about the distribution of the 3. Form the ratio
2. Does there exist a general principle that yet such a departure may have little maximum likelihood for θ in Ω0
population being sampled, how well will practical significance. λ ( x1,..., xn ) =
can be used to obtain best or good test maximum likelihood for θ in Ωa
the test procedure work when the
procedures?
assumptions are violated? Reject H0 when this ratio is small.