Академический Документы
Профессиональный Документы
Культура Документы
Hypothesis Testing
• Test of hypothesis - Test whether a
Hypothesis Testing population parameter is less than, equal
to, or greater than a specified value.
• Remember an inference without a
measure of reliability is little more than a
guess.
Example Example
• A contract between a city and a road • Suppose building specifications for a city require
construction company called for an asphalt road that the average breaking strength of residential
with an average thickness of 10 inches. The sewer pipe to be more than 2,400 lbs per foot of
county thought that the construction company length. To sell pipe in the city a manufacturer
must demonstrate that its product meets the
had defrauded the city by making the road less specifications.
than 10 inches thick. The city then took 100
• Here we are interested not so much in the value
core samples 1 inch in diameter. The mean of than we are in testing a hypothesis about its
thickness of the samples was 9.5 inches. Did value. In other words, we are interested in
fraud occur? whether the mean breaking strength of the pipe
exceeds 2,400 lbs per foot.
1
Chapter 9
Hypotheses Statements
Hypotheses Statements
Examples
• Two hypotheses: • For pipe example:
– Null hypothesis (H0) - status quo to the party – (H0) 2 , 400 (the pipe does not meet
performing the experiment. The hypothesis will not specifications)
be rejected unless the data provide convincing
evidence that it is false. – (Ha) 2 , 4 0 0 (the pipe meets specifications)
– Alternative or research hypothesis (Ha) - which will • How can the city decide when enough evidence
be accepted only if the data provide convincing exists to conclude the pipe meets specifications?
evidence of its truth. – When the sample mean x convincingly indicates that
the population mean exceeds 2,400 lbs per foot.
Evidence Example
• “Convincing” evidence in favor of the alternative • The test statistic is:
exists when the value of x exceeds 2,400 by x x 2 , 400
z
an amount that cannot be readily attributed to x
n
sampling variability.
• We need to compute a “test statistic” which will
• A value of z=1 means that x-bar is 1
standard error above 2 ,4 0 0
be the z-value that measures the distance
between the values of x and the value of
specified in the null hypothesis. • A value of z=1.5 means that x-bar is 1.5
standard errors above 2 ,4 0 0
2
Chapter 9
11
3
Chapter 9
4
Chapter 9
Example P-Values
• A tire company guarantees that a particular tire • For all the hypothesis testing examples so far,
has a mean useful lifetime of 42,000 miles or the value of and thus the rejection region are
more. A consumer testing agency, wishing to selected prior to conducting the test.
verify this claim, observed n=10 tires on a test
wheel that simulated normal road conditions.
The lifetimes (in thousands of miles) were as • A second method is to report the extent to which
follows: 42, 36, 46, 43, 41, 35, 43, 45, 40, 39. the statistic disagrees with the null hypothesis
Use these data to determine whether there is and letting the reader decide whether to reject
sufficient evidence to contradict the the null hypothesis. This measure of
manufacturers claim at the .05 level. disagreement is called the p-value.
5
Chapter 9
z z
z z
2
6
Chapter 9
Example
• Where • A rating service for local radio stations claims
x that 40% of the prime drive-time audience listens
p
n to WXIX. You are the advertising manager of a
competing radio station and you doubt that
p 0q WXIX’s market share is that large. Formulate
p 0
n and test the appropriate hypothesis at the 5%
significance level if 380 people out of 1,000
surveyed say that they listen to WXIX. What is
the p-value?
7
Chapter 9
Large Sample Hypothesis Test for Small Sample Confidence Interval for
(1 - 2) - Independent Sampling (1 - 2) - Independent Sampling
• One-tailed Test: • Both populations must be approximately
– H0: (1 - 2) = D0 (or 0) vs Ha: (1 - 2) > D0 (or 0)
– H0: (1 - 2) = D0 (or 0) vs Ha: (1 - 2) < D0 (or 0)
normally distributed with equal variances (these
variances are usually unknown):
• Two-tailed Test:
– H0: (1 - 2) = D0 (or 0) vs Ha: (1 - 2) not equal to D0 (or 0)
( x1 x 2 ) t ( x1 x 2 )
2
• Test statistic:
x1 x2 D0 1 1
zcalc ( x1 x 2 ) t s 2p
x x 2 n1 n2
1 2
2
2
2
n 1 1 s 12 n 2 1 s 22
1
2 s
n1 n2 2
(X X ) p
1 2
n1 n2
8
Chapter 9
1 1
s 2p
n1 n2 performing hypothesis tests and calculating
confidence intervals.
s 2p
n1 1s12 n2 1s22
n1 n2 2
9
Chapter 9
10