0% found this document useful (0 votes)

71 views32 pages

Appendix C Mathematical Statistics 2015

The document outlines the topics covered in a course on fundamental mathematical statistics. It is divided into three parts. Part one covers populations, parameters, and random sampling, finite sample properties of estimators, and asymptotic properties of estimators. Part two discusses general approaches to parameter estimation and interval estimation and confidence intervals. Part three covers hypothesis testing and remarks on notation. The first section defines key concepts such as populations, parameters, and random sampling and gives examples to illustrate estimation and hypothesis testing.

Uploaded by

Sydney

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

71 views32 pages

Appendix C Mathematical Statistics 2015

Uploaded by

Sydney

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 32

Outline:

Fundamentals of Mathematical Statistics

Part One
I. Populations, Parameters, and Random Sampling
II. Finite Sample Properties of Estimators
Fundamentals of Mathematical Statistics III. Asymptotic or Large Sample Properties of Estimators

Part Two
IV. General Approaches to Parameter Estimation
V. Interval Estimation and Confidence Intervals

Read Wooldridge, Appendix C: Part Three
VI. Hypothesis Testing
VII. Remarks on Notation

I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
2
Fundamentals of Mathematical Statistics: Part One . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat

I. Populations, Parameters, and Populations, Parameters, and
Random Sampling Random Sampling
• By “learning”, we can mean several things.
• Population refers to any well‐defined group of subjects. – Most important are estimation and hypothesis testing.

Example:
• Statistical inference involves learning something about the • Suppose our interest is to find the average percentage increase in wage given an
additional year of education.
population from a sample. – Population: obtain wage and education of 33 million working people
– Sample: obtain data on a subset of the population.

• Parameters are constants that determine the directions and Example: Results:

strengths of relationship among variables. o the return to education is 7.5%
‐ example of point estimate.
o the return to education is between 5.6% and 9.4%
‐ example of interval estimates.
o Does education affect wage?
– example of hypothesis testing.

Example: a random sample from normal distribution.

• If Y1, …, Yn are independent random variables with a normal distribution with mean
and  variance 2, then {Y1, …, Yn} is a random sample from the Normal(,2)
population.

Sampling Sampling
Example: random sample from Bernoulli distribution.
Example: working population
• If Y1, ..., Yn are independent random variables, each is distributed as
• We may obtain a sample of 100 families. Bernoulli () so that
– Note that the data we observe will differ for each different sample. A P(Yi=1) = 
sample provides a set of numbers, say, {y1, …, yn}. P(Yi=0) = 1 ‐ 
then, {Y1, ..., Yn} constitutes a random sample from the Bernoulli ()
distribution.

• Note that
Yi = 1 if passenger i show up
Yi = 0 otherwise

• Suppose {Y1, …, Yn} is a random sample from a population distribution
that depends on an unknown parameter .
• A “finite sample” implies a sample of any size, no matter how large or – An estimator of  is the rule that assigns each possible outcome of the
small. sample.
– Small sample properties.
– A rule is specified before any sampling is carried out.

• An estimator W of a parameter  can be expressed as

• Asymptotic properties have to do with the behavior of estimators as the
sample size grows without bound.
W = h(Y1, …, Yn}

for some known function h.

• When [particular set of values, say {y1, …, yn}, is plugged into the function h, we
obtain the estimate of .

A. Unbiasedness A. Unbiasedness
Estimators and Estimates: B. Variance B. Variance
C. Efficiency Estimators and Estimates C. Efficiency
sampling distribution
Example:
• The distribution of an estimator is called the sampling distribution.
– It describes the likelihood of various outcomes of W across different
• Let {Y1, …, Yn} be a random sample from the population with mean .
The natural estimator of  is the average of the random sample:
random samples.

• The entire sampling distribution of W1 can be obtained given the
probability distribution of W1 and outcomes. Note that Y‐bar is called the sample average.

• Unlike in Appendix A, we define the sample average of a set of numbers
as a descriptive statistic.

• For actual data outcomes, y1, …, yn, the estimate is the average in the
sample
…

Unbiased Estimator: Definition
• Example:
An estimator W of  is unbiased if

E(W) = 
for all possible values of W

• Estimator:
– Intuitively, if the estimator is unbiased, then its probability distribution has an
expected value equal to the parameter it is supposed to be estimating.
• Estimate: = 6.0
– Our estimate of the average city unemployment rate in the U.S. is 6.0%.
• Unbiasedness does not mean that the estimate from a particular
Notes sample is equal to , or even very close to .
1) Each sample results in a different estimate. • If we could indefinitely draw random samples on Y from the population
2) The rule for obtaining the estimate is the same. – then average these estimates over all random samples will obtain .

A. Unbiasedness A. Unbiasedness
B. Variance B. Variance
Unbiasedness C. Efficiency Unbiasedness C. Efficiency

Bias of an Estimator: Definition

If W is an estimator of , its bias is defined as • Show: the sample average is an unbiased estimator of the population
mean .
Bias() = E(W) – 
An estimator has a positive bias if E(W) –  >0. 1 n  1  n  1 n 
E (Y )  E   Yi   E   Yi     E (Yi ) 
•
 i 1 
n n  i 1  n  i 1 
1 n  1
•
–
The unbiasedness of an estimator and the size of bias depend on
the distribution of Y and
    ( n )  
n  i 1  n
– the function h

• We cannot control the distribution of Y, but we could choose the choice
of the rule h.

Weaknesses:
(1) Some very good estimators are not unbiased. • The variance of an estimator is the measure of the dispersion in the
distribution.
(2) Unbiased estimators could be quite poor estimators. It is often called sampling variance.
• Example: the variance of sample average from a population.
Example:
Let W = Y1 (from a random sample of size n, we discard all of the observations except
the first)
E(Y1) = 

• Unbiasedness ensures that the probability distribution of an estimator has a • Summary:

mean value equal to the parameter it is supposed to be estimating. If {Y1, …, Yn} is a random sample from a population with mean  and variance
• Variance shows how spread out the distribution of an estimator. 2, then
• has the same mean as the population
• Its sampling variance equals the population variance 2 over the sample size.
(2/n)

A. Unbiasedness A. Unbiasedness
The Sampling Variance of B. Variance The Sampling Variance of B. Variance
C. Efficiency C. Efficiency
Estimators Estimators

Suppose W1 and W2 both are unbiased This implies that the probability that W1 is

• Define the estimator as estimators of , but W1 is more tightly centered greater than any distance from  is less than
the probability W2 is greater than the same
about . (See graph!) distance from .

which is usually called the sample variance.

• Show that the sample variance is an unbiased estimator of 2.
E(S2) = 2

Example: Example:
For a random sample with mean  and variance 2. From simulation in Table C.1.
Y1 is the estimator, the first observation drawn.
20 random samples of size 10
(n=10) generated from the normal
Estimator Estimator Y1 distribution with =2 and 2=1
Mean is unbiased Y1 is unbiased
E( ) =  E(Y1) =   y1 ranges from (‐0.64‐4.27);
Variance Var( )= 2/n Var(Y1)= 2 mean = 1.89
 ranges from (1.16‐2.58)
If the sample size n=10; this implies Var(Y1) is ten times larger than mean = 1.96
Var( )
 Which estimator is better?

Relative Efficiency A. Unbiasedness A. Unbiasedness

B. Variance
C. Efficiency Efficiency B. Variance
C. Efficiency

Example:
Relative Efficiency: If W1 and W2 are two unbiased estimators of , then • For estimating the population mean ,
W1 is efficient relative to W2 when Var(W1)Var(W2) for all . Var( ) < Var(Y1) for any value of 2.

• The estimator is efficient relative to Y1 for estimating .

• In a certain class of estimators, we can show that the sample average has the
smallest variance.
Example:

Example:y1
Example:
Show that has the smallest variance among all unbiased estimators that are
also linear functions of Y1, Y2, …, Yn.
– The assumptions are that Yi have common mean and variance, and that they are
pairwise uncorrelated.

Efficiency B. Variance
C. Efficiency Efficiency
B. Variance
C. Efficiency

• If we do not restrict our attention to unbiased estimators, then
comparing variances is meaningless • A measure when comparing estimators that are not necessarily
unbiased:
– Mean squared error (MSE)
Example: In estimating the population mean , we use trivial estimator equal
to zero • If W is an estimator of , then
– mean equal to zero E(0) = 0 MSE(W) = E(W‐)2
– Variance equal to zero: Var(0) = 0 = E[W‐E(W) +E(W)‐]2
– bias of this estimator equal ‐  Bias(0) = ‐  = Var(W) + [bias(W)]2
Bias(0) = E(0) ‐  = ‐ 
• The MSE measures how far, on average, the estimator is away from . It
• So this trivial estimator is a very poor estimator when the bias of the depends on the variance and bias.
estimator or  is large.

A. Unbiasedness A. Unbiasedness

Problem C.1
B. Variance
C. Efficiency Problem C.1 continue.. B. Variance
C. Efficiency

C.1 Let Y1, Y2, Y3, and Y4 be independent, identically distributed (ii) Now, consider a different estimator of :

random variables from a population with mean  and
1 1 1 1
variance 2. Let W  Y1  Y2  Y3  Y4
8 8 4 2
1
Y  (Y1 + Y2 + Y3 + Y4) This is an example of a weighted average of the Yi. Show
4 that W is also an unbiased estimator of . Find the
denote the average of these four random variables. variance of W. [ans.]

(i) What are the expected value and variance of in terms of  (iii) Based on your answers to parts (i) and (ii), which

and 2? [ans.] estimator of  do you prefer, or W? [ans.]

I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
II. Finite Sample Properties of Estimators Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 27 II. Finite Sample Properties of Estimators Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 28
Problem C.1 (i) continue… Problem C.1 (ii)
(i) This is just a special case of what we covered in the text, with n = 4:
(ii)
• E( ) = µ and Var( ) = 2/4.
• W is unbiased.
• Expected value of  = E(W) = E(Y1)/8 + E(Y2)/8 + E(Y3)/4 + E(Y4)/2
= µ[(1/8) + (1/8) + (1/4) + (1/2)]
1 n  1  n  1 n  1 n
 1
E (Y )  E   Yi   E   Yi     E (Yi )   
 n i 1  n  i 1  n  i 1  n
    ( n )  
i 1  n
= µ(1 + 1 + 2 + 4)/8 = µ,
Note that Yi are independent.
• Variance of = 2/4.
• Find variance of W
Var(W) = Var(Y1)/64 + Var(Y2)/64 + Var(Y3)/16 + Var(Y4)/4
= 2[(1/64) + (1/64) + (4/64) + (16/64)]
= 2(22/64)
= 2(11/32).
29
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
30
II. Finite Sample Properties of Estimators Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat II. Finite Sample Properties of Estimators Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat

III. Asymptotic or Large Sample A. Consistency
Problem C.1 (iii) Properties of Estimators
B. Asy. Normality

• For estimating a population mean 
• (iii) – One notable feature of Y1 is that it has the same variance for any sample
size
Var(W) = 11/32 – improves in the sense that its variance gets smaller as n gets larger.
Var( ) = 8/32 = ¼ • Y1 does not improve in this case

• We can rule out silly estimators by studying the asymptotic or large
• Because 11/32 > 8/32 = 1/4, Var(W) > Var( ) for any sample properties of estimators (n).
2 > 0, so is preferred to W because each is unbiased.
• How large is “large” sample?
– This depends on the underlying population distribution.
– Note that large sample approximations have been known to work well for
sample sizes as small as 20 observations (n=20).

Consistency B. Asy. Normality Consistency B. Asy. Normality

1. The distribution Wn becomes 3. When Wn is consistent, we say that  is

more and more concentrated about  the probability limit of Wn, written as
as sample size increases (n). plim(Wn) = 
• Consistency concerns how far the estimator is likely to be from the
parameter it is supposed to be estimating 2. For larger sample sizes, Wn is less and 4. The conclusion that Yn is consistent of  is
– as sample size increases indefinitely. less likely to be very far from . known as the law of large numbers (LLN)

• Definition: Consistency
Let Wn be an estimator of  based on Y1, …, Yn of sample size n. Then, Wn is a
consistent estimator if for for every >0

P(Wn‐ ) >   0 as n  

• Note that we index the estimator by the sample size, n, in stating this
definition.

I. Random Sampling II. Finite Sample III. Asymptotic Sample I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
33 III. Asymptotic or Large Sample Properties of Estimators Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 34
Fundamentals of Mathematical Statistics: Part One . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat

A. Consistency A. Consistency
Consistency B. Asy. Normality
Law of Large Number B. Asy. Normality

• Unbiased estimators are not necessarily consistent, but those whose variances
shrink to zero as sample size increases are consistent.
Definition: LLN
Let Y1, …, Yn be i.i.d. random variables with mean . Then
• Formally, if Wn is an estimator of  and Var(Wn)0 as n  , then plim(Wn)=.
plim( n) = 

• Example: the average of a random sample drawn with mean  and variance 2
Intuitively,
– the sample average is unbiased
the LLN says that if we are interested in finding population average , we get
2 arbitrarily close to  by choosing a sufficiently large sample.
Var (Yn ) 
n

– Thus, Var( n)  0 as n ‐> 

n is a consistent estimator of 

Consistency B. Asy. Normality

Property PLIM.1
Let  be a parameter and define a new parameter =g() for some continuous Property PLIM.2
function g( ). Suppose plim(Wn)= . Define an estimator of  as Gn=g(Wn). Then

plim(Gn) = .
If plim(Tn)= and plim(Un)=, then
Alternatively,
1) plim(Tn+Un) =  + 
plim[g(Wn)] = g[plim(Wn)]
2) plim(TnUn) = 
for some continuous function, g().
3) plim(Tn/Un) = / provided that   0
• What is a continuous function?
– Note a continuous function is a “function that can be graphed without lifting your pencil from the
paper”.

Examples:
• g()= a + b
• g()= 2
• g() = 1/
• g() = exp()

I. Random Sampling II. Finite Sample III. Asymptotic Sample I. Random Sampling II. Finite Sample III. Asymptotic Sample
37 38
Fundamentals of Mathematical Statistics: Part One . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat Fundamentals of Mathematical Statistics: Part One . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat

A. Consistency A. Consistency
Consistency B. Asy. Normality
Consistency B. Asy. Normality

• Example:
∑ Example:
(1) E
Estimating standard deviation from a population with mean  and variance
∑ 2
(2) Yn* E(Y*)
Given sample variance

• As n  , and X* are consistent estimators of .

(1) plim( ) =  Var( ) = 0 as n 
– Sample variance is unbiased estimator for 2.
∑ – Sn2 is also a consistent estimator for 2.
(2) plim(Y*) =  plim( = plim( plim( )= 1* = 

• Sample standard deviation
Var( ) = 0 as n  – Sn is not an unbiased estimator of . Why?
– Sn is a consistent estimator of .
• Y* is also a consistent estimator since Y* approaches the value of the plim S n  plim S n   2  
parameter  as sample size gets larger and larger.

Example: Central Limit Theorem

Yi annual earnings with a high school education
(population mean = Y)
Zi annual earnings with a college education
Consistency is a property of point estimators, so is unbiasedness.
(population mean = Z)
• Let {Y1,…, Yn} and {Z1,…, Zn} be a random sample of size n from a population of workers, and
we want to estimate the percentage difference in annual earnings between two groups, Consistency and distribution
which is • Consistency does not tell us about the shape of that distribution for a given
  100  (  Z  Y ) / Y sample size.
• Most econometric estimators have distributions that are well approximated by
a normal distribution for large samples (n).
• Due to the facts that
plim( Z n )   Z plim(Yn )  Y

• It follows from PLIM.1 and PLIM.2 that
Gn  100  ( Z n  Yn ) / Yn

• Gn is a consistent estimator of . It is just the percentage difference between ̅ n and n.
plim(Gn )  

A. Consistency A. Consistency
Asymptotic Normality B. Asy. Normality
Central Limit Theorem (CLT) B. Asy. Normality

Definition: Asymptotic Normality
• Let {Zn: n=1, …, n} be a sequence of random variables such that for all numbers z, • Definition: CLT
Yi  d(, 2) implies that {Y1, …, Yn} be a
P(Zn z)  (z) as n  , If Yi  d(, 2), then random sample with mean  and variance
2.

where (z) is the standard normal cumulative distribution function (cdf) The variable Zn is the standardized

version of n: we have subtracted off
E( n)= and divided by sd( n)=/ .

has an asymptotic standard normal distribution. “a” stands for “asymptotically”

or “approximately”.
cdf for Zn
(z)  as n

• Intuitively,
The central limit theorem (CLT) suggests that the
average from a random sample for any population,
when standardized, has an asymptotic standard normal distribution.
│z
ZN

• Intuitively,
This property means that the cdf for Zn gets closer and closer to the cdf of the
standard normal distribution as the sample size n gets large.

The exact distributions of W1 = [(n‐1)/n] and

W2 = /2.

(i) Show that W1 and W2 are both biased estimators of  and find

are not the same as (1) , but the difference is often small enough to be the biases. What happens to the biases as n  ? Comment
ignored for large n. on any important differences in bias for the two estimators as
the sample size gets large. [ans.]

A. Consistency
B. Asy. Normality
Problem C.3 continue … Problem C.3 (i)
(i)
(ii) Find the probability limits of W1 and W2. {Hint: Use
• E(W1) = [(n – 1)/n]E( ) = [(n – 1)/n]µ,
properties PLIM.1 and PLIM.2; for W1, note that
Bias(W1) = [(n – 1)/n]µ – µ = –µ/n.
plim[(n‐1)/n] = 1.} Which estimator is consistent? As n , Bias(W1)  0
[ans.]
• Similarly,
(iii) Find Var(W1) and Var(W2). [ans.] E(W2) = E( )/2 = µ/2,
Bias(W2) = µ/2 – µ = –µ/2.
As n , Bias(W1) = –µ/2
(iv) Argue that W1 is a better estimator than if  is
“close” to zero. [ans.] • The bias in W1 tends to zero as n  , while the bias in W2 is –µ/2 for
all n.  This is an important difference.
(Consider both bias and variance.)

• Because plim(W1) = µ and plim(W2) = µ/2, W1 is • Var(W2) = Var( )/4 = 2/(4n).

consistent whereas W2 is inconsistent.

A. Moments
IV. General Approaches to B. Max Likelihood
Problem C.3 (iv) Parameter Estimation
C. Least Squares

(iv)
• Because is unbiased, its mean squared error is simply its variance.   • We have learned finite and asymptotic properties for estimation –
MSE( ) = Var( ) + [Bias( )]2 unbiasedness, consistency and efficiency.
= 2/n
• On the other hand, Question:
MSE(W1) = Var(W1) + [Bias(W1)]2 Are they general approaches that produce estimators with good properties?
= [(n – 1)2/n3]2 + µ2/n2.
Let µ = 0,
MSE(W1) = Var(W1) = [(n – 1)2/n2](2/n) • Given a parameter  appearing in a population distribution, there are
usually many ways to obtain unbiased and consistent estimator of .
Thus, MSE(W1) < MSE( ) or
Var(W1) < Var( ) • There are three methods
[(n – 1)2/n2](2/n) < 2/n because (n – 1)/n < 1.   – Method of Moments
– Method of Maximum Likelihood
• Therefore, MSE(W1) is smaller than Var( ) for µ close to zero. – Method of Least Squares

• For large n, the difference between the two estimators is trivial.

Example: Population covariance
• The basis of the method of moments proceeds as follows. • The population covariance between two random variables X and Y is
XY = E(X‐X)(Y‐Y)
– The parameter  is shown to be related to some function of some expected
value in the distribution of Y, usually E(Y) and E(Y2).
• The method of moment suggests the following estimator,
1 n
S XY   ( X i  X )(Yi  Y )
n i 1

Example: Population mean 1)This is a consistent estimator of XY.

• Suppose  is a function of ; i.e., =g() 2) However, this is a biased estimator.

• Given the sample average is an unbiased and consistent estimator of , Example: Sample covariance

it is natural to replace  for . Thus, g( ) is the estimator .
– In addition, the estimator g( ) is a consistent estimator of . If g () is linear • The sample covariance is
function of , then g( ) is an unbiased estimator of .

• Why is this a method of moments?
– Here we replace the population moment  with the sample average . 1) It can be shown that this is an unbiased estimator of XY.
2) This is a consistent estimator of XY.

A. Moments A. Moments
B. Max Likelihood B. Max Likelihood
Method of Moments C. Least Squares Maximum Likelihood C. Least Squares

Example: Population correlation • Maximum likelihood Estimator (MLE) 1. In the discrete case, this is

P(Y1=y1, Y2=y2,…, Yn=yn)
XY = XY/(XY) = P(Y1=y1)P(Y2=y2) … P(Yn=yn)
• Let {Y1, …, Yn} be a random sample from the 2. The joint distribution of {Y1, …, Yn}
population distribution f(y;). can be written as the product of
• The method of moments suggests estimating XY as the densities: f(Y1;)f(Y2,)  f(Yn,).

• The likelihood function, which is a random variable, can be defined as

L(; Y1, …, Yn) = f(Y1;)f(Y2;)  f(Yn;)

This is called the sample correlation coefficient. • It is easier to work with the log‐likelihood function
Facts:
1) It is obtained by taking the natural
log of the likelihood function.

2) The log of the product is the sum

Notes: of the logs.
(1) RXY is a consistent estimator of XY.
(Why?): SXY, SX and SY are consistent.
• Then, the maximum likelihood estimator of , call it W, is the value of  that maximizes the
(2) RXY is not an unbiased estimator of XY. (Why?) likelihood function.
First: SX and SY are not unbiased estimators. – Intuitively, out of the possible values for , the value that makes the likelihood that the
Second: RXY is a ratio of estimators, so it would not be unbiased. observed values are largest should be chosen.

• Least squares estimators – a third kind of estimator.
• Properties: • The sample mean, , is a least square estimator of the population mean .
1) MLE is usually consistent and sometimes unbiased. – It can be shown that the value of m which make the sum of squared deviations as small
as possible is m =
2) MLE is the generally the most asymptotically efficient estimator (when
the population model f(y;) is correctly specified).
– MLE has the smallest variance among all unbiased estimators of .
– MLE is the minimum variance unbiased estimator.

Properties:
1) LSE is consistent and unbiased.

2) LSE is the generally the most efficient estimator in finite and large samples.

– LSE has the smallest variance among all linear unbiased estimators of .

A. Nature
A. Moments
B. CI N(0,1)
Methods of Moments, Least B. Max Likelihood V. Interval Estimation and C. Rule of Thumb
C. Least Squares
Squares, Maximum Likelihood Confidence Intervals D. Asymptotic CI

• A point estimate is the researcher’s best guess at the population value,
• The principles of least squares, method of moments, and but it provides no information how close the estimate is likely to be to the
maximum likelihood often result in the same estimator. population parameter.

Summary Unbiased Consistent Efficiency Example:

• On the basis of a random sample of workers,
Moments unbiased consistent efficient
– a researcher reports that job training grants increase hourly wage by 6.4%
Least Squares unbiased consistent efficient – We cannot know how close an estimate is for a particular sample because we
Maximum usually asymptotically do not know the population value.
consistent
Likelihood consistent efficient

• Interval estimation comes in when we make statement involving • The standardized version of has a standard normal distribution.

probabilities.
 Y  
– One way of assessing the uncertainly in an estimator – sampling standard deviation. • Rewrite as P  1.96   1.96   .95
 / n 

• Interval estimation uses information on the point estimate and the standard • The random interval is P(Y 1.96 / n <   Y  1.96 / n )  0.95

deviation by constructing Confidence interval.
(Y  1.96 / n , Y  1.96 / n )
– It shows how the population value is likely to lie in relation to the estimate.

1) The probability that the random interval contains the population mean  is .95 or 95%
Concept of Interval estimation: 2) This information allows us to construct an interval estimate of 
‐ by plugging the sample outcome of the average, and  =1 into the random interval .
• Assume {Y1, …, Yn} is a random sample from the Normal(, 2) population.
Suppose that the variance 2 is known (or 2=1). P(Y 1.96 / n <   Y  1.96 / n )  0.95
– The sample average has a normal distribution with mean  and variance
2 /n; i.e., • It is called a 95% confidence interval.

–  Normal(, 2 /n). • (Y  1.96 / n , Y  1.96 / n )

A shorthand notation is random interval
y  1.96 / n confidence interval

A. Nature A. Nature
B. CI N(0,1) B. CI N(0,1)
The Nature of Interval Estimation
The Nature of Interval Estimation C. Rule of Thumb C. Rule of Thumb
D. Asymptotic CI (Y  1.96 / n , Y  1.96 / n ) D. Asymptotic CI

Example:
• Given the sample data {y1, y2, …, yn} are observed. We can find .
Suppose n= 16, =7.3, =1 Correct Interpretation:
A random interval contains  with probability 0.95.
• The 95% confidence interval for  is
7 .3  1 .96 / 16 = 7.3 .49
Incorrect interpretation:
• We can write in an interval form as [6.81, 7.89]. The probability that  is in the interval is 0.95.
• The meaning of a confidence is more difficult to understand. • since  is unknown and it either is or is not in the
We mean that the random interval contains  with probability .95 interval.
– There is a 95% chance that random interval contains .

• Random interval is an example of interval estimator.
– since endpoints change with different samples.

Example:
• Table C.2 contains calculations for 20
random samples • Suppose the variance is 2 and known.  The 95% confidence interval is
• Assume Normal(2,1) distribution with
sample size n=10.
• Interval estimates of  are .62.
• In practice, we rarely know the population variance 2.
Results: • To allow for unknown , we can use an estimate:
1)  The interval changes with each random
sample.
2)  19 of the 20 intervals contain the
population value of . • However, the random interval no longer contains  with probability .95 because
3) Only for replication number 19, =2 is not the constant  has been replaced with the random variable S.
in the confidence interval.
Y  1 .96 ( S / n )
4) 95% of the samples result in a confidence
interval that contain .

I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
IV. Parameter Estimation V. Interval Estimation & Confidence Interval
Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
65 66
V. Interval Estimation and Confidence Intervals
Fundamentals of Mathematical Statistics: Part Two . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat

A. Nature A. Nature
CIs for the Mean from a Normally B. CI N(0,1) CIs for the Mean from a Normally B. CI N(0,1)
C. Rule of Thumb C. Rule of Thumb
Distributed Population D. Asymptotic CI Distributed Population D. Asymptotic CI

• We use t distribution, rather standard normal distribution.
Example:
• Let n=20
where S is the sample standard deviation of the random sample {Y1, …, Yn}. df = n‐1 = 19
c = 2.093 (See Table G.2 in Appendix
– Note that G)

• The 95% confidence interval is
• To construct a 95% confidence interval using t
distribution, let c be the 97.5th percentile in
y  2.093( s / 20 )
the tn‐1 distribution. P(-c<t<c) =.95
and s are the values obtained
• Once the critical value c is chosen, the random from the sample.
interval contains .
Table G.2 in Appendix G.
[Y  c  S / n , Y  c  S / n ]
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
67 Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
68
V. Interval Estimation and Confidence Intervals V. Interval Estimation and Confidence Intervals
A. Nature A. Nature
B. CI N(0,1) Example C.2 Effect of Job Training Grants B. CI N(0,1)
CIs for the Mean from a Normally C. Rule of Thumb on Worker Productivity C. Rule of Thumb
Distributed Population D. Asymptotic CI D. Asymptotic CI

• A sample of firms receiving job training grants
in 1988.
• More generally, let c be the 100(1‐/2) percentile in the tn‐1 distribution.
 Scrap rates – number of items per 100
produced that are not useable and need
• A 100(1‐ )% confidence interval is to be scrapped.
 The change in scrap rates has a normal
c/2 – known after choosing  and degree of freedom n‐1. distribution.
• n=20, = ‐1.15
• Recall that sd(Y )  / n
se( ) = s/ = .54 (Note that s=2.41)

• s/ is the point estimate of sd( or the standard error of . • A 95% confidence interval for the mean

se(Y )  s / n change in scrap rate  is
 2.093se( )
• A 100(1‐)% confidence interval can be written as [‐2.28, ‐0.02]

• With 95% confidence, the average change in
scrap rates in the population is not zero!
• The notion of the standard error of an estimate plays an important role in econometrics.
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
69 Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
70
V. Interval Estimation and Confidence Intervals V. Interval Estimation and Confidence Intervals

A. Nature A. Nature
B. CI N(0,1) A Simple Rule of Thumb for a B. CI N(0,1)
Example C.2 Effect of Job Training C. Rule of Thumb C. Rule of Thumb
D. Asymptotic CI 95% Confidence Interval D. Asymptotic CI
Grants on Worker Productivity
• A Rule of Thumb for an approximate 95% confidence interval is
• The analysis above has some potentially serious flaws.

• It assumes that any systematic reduction in scrap rates is due to the job
training grants. 1) It is slightly too big for large sample sizes
– Many things (variables) can happen over the course of the year to change 2) It is slightly too small for small sample sizes.
worker productivity!

• Note that t distribution approaches the standard normal distribution as the
degrees of freedom gets large.

• In particular,
=.05, c/2  1.96 as n
(see graph)

• For some applications, the population is nonnormal. • Matched pairs analysis – each person in a pair interviewed for the same job.

– In some cases, the nonnormal population has no standard distribution. B Probability that the black person is offered a job
• This does not matter as long as sample sizes are sufficiently large for the central limit w Probability that the white person is offered a job
theorem to give a good approximation of the distribution of the sample average .
• A large sample size has a nice feature since it results in small confidence intervals. This • We are interested in the difference B ‐ W
is because standard error for
Bi=1 If the black person gets a job offer from employer i
– [se( )] shrinks to zero as the sample size grows.
Wi=1 If the white person gets a job offer from employer i

• Unbiased estimators of B and W are and the fractions of interviews for which

blacks and whites were offered jobs.
• For large n, an approximate 95% confidence interval is

• A new variable Yi =-1
if the black did not get the job, but
the white did
where 1.96 is the 97.5th percentile in the standard normal distribution. Yi = Bi – Wi
Yi =0 if both did or did not get the job

• Note that the standard normal distribution is used in place of t distribution since we deal • Yi can take these values, then,
if the white did not get the job, but
Yi =1
with asymptotics.   E ( B i )  E (W i )   B   W the black did
– as n increases without bound, the t distribution approaches standard normal distribution.

A. Nature A. Nature
Example C.3 Race B. CI N(0,1) B. CI N(0,1)

Discrimination in Hiring
C. Rule of Thumb
D. Asymptotic CI Problem C.7 C. Rule of Thumb
D. Asymptotic CI

C.7 The new management at a bakery
• Sample size n=241 b  .224 and w  .357, so y  .224  .357  .133 claims that workers are now more
obs Wb Wa D=Wa-Wb

1 8.3 9.25 0.95

productive than they were under old 2 9.4 9 -0.4
– 22.4% of black were offered jobs, while 35.7% of white were offered jobs management, which is why wages have 3 9 9.25 0.25
– This is prima facie evidence of discrimination! “generally increased.” Let Wib be Worker 4 10.5 10 -0.5
i’s wage under the old management and
• Sample standard deviation: s=0.482 let Wia be Worker i’s wage after the 5 11.4 12 0.6

change. The difference is 6 8.75 9.5 0.75

• Find an approximate 95% confidence interval for  7 10 10.25 0.25

Di = Wia ‐ Wib. 8 9.5 9.5 0
Assume that the Di are a random sample
• A 95% CI for  = B ‐ w is ‐.133  1.96(.482/(241)½ 9 10.8 11.5 0.7
from a Normal(, 2) distribution. 10 12.55 13.1 0.55
‐.133 .031
11 12 11.5 -0.5
[‐.164, ‐.102]
(i) Using the following data on 15 workers, 12 8.65 9 0.35

• A 99% CI for  = B ‐ w is ‐.133  2.58(.482/(241)½

construct an exact 95% confidence 13 7.75 7.75 0

interval for . [ans.] 14 11.25 11.5 0.25

[‐.213, ‐.053]
15 12.65 13 0.35

mean 10.16667 10.40667 0.24

• We are very confident that the population difference is not zero!

2 9.4 9 -0.4 Mean 10.16667 10.40667 0.24

• The average increase in wage is 3 9 9.25 0.25
Median 10 10 0.25
̅ = .24, or 24 cents. 4 10.5 10 -0.5
Maximum 12.65 13.1 0.95
5 11.4 12 0.6
Minimum 7.75 7.75 -0.5
6 8.75 9.5 0.75
Std. Dev. 1.569084 1.595291 0.450872
• The sample standard deviation is 7 10 10.25 0.25
Skewness 0.175376 0.290842 -0.34947
about s = .451 8 9.5 9.5 0
Kurtosis 1.810807 2.022774 2.161199
9 10.8 11.5 0.7
n = 15, 10 12.55 13.1 0.55

se( ̅ ) = .1164. 11 12 11.5 -0.5

Jarque-Bera 0.960754 0.80833 0.745067

12 8.65 9 0.35 Probability 0.61855 0.667534 0.688986

13 7.75 7.75 0

14 11.25 11.5 0.25 Sum 152.5 156.1 3.6

15 12.65 13 0.35 Sum Sq. Dev. 34.46833 35.62933 2.846
mean 10.16667 10.40667 0.24

Observations 15 15 15

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
C. Asymptotic Fundamentals of Hypothesis Testing C. Asymptotic
VI. Hypothesis Testing D. P-Value
E. CI & HT
D. P-Value
E. CI & HT

• Suppose the election results are as follows:
• We have reviewed how to evaluate point estimators and to – Candidate A=42% and
construct confidence intervals. – Candidate B=58% of the popular vote
– Sometimes the question we are interested in has a definite yes or • Candidate A argued that the election was rigged.
no answer. Consulting agency: a sample of 100 voters. It was found that 53% voted for Candidate A.
1) Does a job training program effectively increase average worker
productivity?
2) Are blacks discriminated against in hiring? Question:
how strong is the sample evidence against the officially reported percentage of
42%?
• Devising methods for answering such questions, using a
sample of data, is known as hypothesis testing. • One way to proceed is to set up a hypothesis test.
Let  be the true proportion of the population voting for Candidate A

• The null hypothesis is
H0:  =.42

• The null hypothesis plays a role similar to that of a defendant.
– A defendant is presumed to be innocent until proven guilty.
– The null hypothesis is presumed to be true until the data strongly suggest • There are two kinds of mistakes:
otherwise.
1)We reject the null hypothesis when it is true – Type I error
Example: We reject H0 when the true proportion of voting for Candidate
• The alternative hypothesis is that the true proportion voting for Candidate A A is in fact 0.42.
is above 0.42.

H1: >.42 2)We “accept” or do not reject the null hypothesis when it is
false – Type II error
• In order to conclude H1 is true and H0 is false, we must prove beyond Example: we “accept” H0, but >.42.
reasonable doubt.
– Observing 43 votes out of a sample of 100 is not enough to overturn the original
result.
• Such an outcome is within the expected sampling variation.

– How about observing 53 votes out of a sample of 100?

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
C. Asymptotic C. Asymptotic
Fundamentals of Hypothesis Testing D. P-Value
E. CI & HT
Fundamentals of Hypothesis Testing D. P-Value
E. CI & HT

• We can compute the probability of making either a Type I or a Type II error.
• Type II error
– Want to minimize the probability of Type II error
• Hypothesis testing requires choosing the significance level, denoted by .
– Alternatively, want to maximize the power of a test.
 = P(Reject H0 H0)
• The power of the test is one minus the probability of a Type II error.
Read: the probability of rejecting null hypothesis, given that H0 is true. Mathematically,

• A significance level is the probability of committing Type I error. () = P(Reject H0) = 1 – P(Type II)

where  the actual value of the parameter.
• Classical hypothesis requires that we specify a significance level for a test.
• We would like the power to equal unity whenever the null hypothesis is
• Common values for  are .10, .05, and .01. false.
– They quantify our tolerance for an error.

• =.05: The researcher is willing to make mistakes (falsely reject H0) 5% of the time.

• Provided that the null hypothesis is true, the critical value c is determined
• In order to test hypothesis, we need to choose a test statistic by the distribution of T and the chosen significance level .
and a critical value.
• All rejection rules depend on the outcome of the test statistic t and the
critical value c.

• The test statistic T is some function of the random sample. • To test hypothesis about the mean  from a

Normal(, 2) is as follows. The null
hypothesis is
• When we compute the statistic for particular outcome, we
H0: = 0,
obtain an outcome of the test statistic, denoted by t.
where 0 is a value we specify. In the
majority of applications, =0.

• The rejection rule depends on the nature of
the alternative hypothesis.

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
Testing Hypotheses about the C. Asymptotic C. Asymptotic

Mean in a Normal Population
D. P-Value
E. CI & HT
Testing Hypotheses about the D. P-Value
E. CI & HT
Mean in a Normal Population

Three alternatives of interest
• For example, for an one‐sided alternative,
• One sided alternative: H1: 0 > 0.
H1:  > 0,
H1:  < 0, • The null hypothesis is effectively
H0:   0.
• Two sided alternative:
H1:   0,
• Here we reject the null hypothesis when the value of
• Here we are interested in any departure from the null hypothesis. sample average, , is sufficiently greater than 0. How?

Testing Hypotheses about the B. HT N(0,1)

C. Asymptotic Example C.4: Effect of Enterprise
B. HT N(0,1)
C. Asymptotic
Mean in a Normal Population D. P-Value
E. CI & HT Zones on Business Investments
D. P-Value
E. CI & HT

• We use standardized version, • Y denotes the percentage change in investment from the year before and year
after a city became an enterprise zone.
• Assume that Y has a Normal(,2) distribution.
• Note that s is used in place of  and se( y )  s / n H0: =0 (Null hypothesis: Enterprise zones have no effect)
• This is called the t statistic. The t statistic measures the distance from to 0 H1: >0 (Alternative hypothesis:They have a positive effect)
relative to the standard error of .

• Under the null hypothesis, the random variable is
• Suppose that we wish to test H0 at the 5% level.  The test statistic is
T  n (Y   0 ) / S
where c is the
• A sample of 36 cities.
T has a tn‐1 distribution. 100(1-)  =0.5; C=1.69 (see Table  G.2)
percentile in a tn-
1 distribution.
 y‐bar=8.2; s=23.9
Example of a one‐tailed test: This is an  t = 2.06
• Choose the significance level =.05.  The critical example of a
value c is chosen so that one-tailed test.
• We conclude that, at the 5% significance level, enterprise zones have an effect on average
P( T > cH0) = .05 investment.
• At 1% significance level, do enterprise zones have an positive effect?
• The rejection rule is t > c

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
C. Asymptotic C. Asymptotic
D. P-Value
E. CI & HT
Testing Hypotheses about the D. P-Value
E. CI & HT
Mean in a Normal Population

• For the null hypothesis and the alternative hypothesis
H0:  ≥ 0,
H1:  < 0.

B • The rejection rule is
a
c t < ‐c
k
This implies that < 0 that are sufficiently far from zero to
reject H0.
U
p

Discrimination in Hiring
D. P-Value
E. CI & HT
Testing Hypotheses about the D. P-Value
E. CI & HT
Mean in a Normal Population
• =B‐W is the difference in the probability that blacks and whites receive job
offers.
 is the population mean of the variable Y=B‐W where B and W are binary variables.
• For the null hypothesis and the alternative hypothesis,
• Testing H0:  = 0,
H0: =0
H1:   0.
H1: <0

• Given n=241, y   .133 se ( y )  .48 / 241  . 031 • The rejection rule is

• The t statistic for testing H0: =0
t> c
t = ‐.133/.031 = ‐4.29

• Critical value = ‐2.58 (one‐sided test; =.005) This gives a two‐tailed test.
• t<‐2.58 There is very strong evidence against H0 in favor of H1.

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
Testing Hypotheses about the C. Asymptotic Testing Hypotheses about the C. Asymptotic
D. P-Value D. P-Value
Mean in a Normal Population E. CI & HT Mean in a Normal Population E. CI & HT

• We have to be careful in obtaining the critical value, c.

• The critical value c (See graph!)
– It is the 100(1‐/2) percentile in a tn‐1 distribution. • Proper language for hypothesis testing:
– If =.05, c is the 97.5th percentile in the tn‐1 distribution. “We fail to reject H0 in favor of H1 at the 5% significance level”

• Incorrect wording:
Example: Let n=22,
“ We accept H0 at the 5% significance level”
• c=2.08, the 97.5th percentile in a t21
distribution.
(See Table G.2)

• Rejection Rule: the absolute value
of t statistic must exceed 2.08.

• Because asymptotic theory is based on n increasing without bound,
• If the sample is large enough, we can invoke central limit – standard normal or t critical values are pretty much the same
theorem.
• Asymptotic theory is based on n increasing without bound • Suggestions:
– For moderate values of n, say between 30 and 60, it is traditional to use the
t distribution.
– For n  120, the choice between two distributions is irrelevant.
• Under the null hypothesis,
T  n (Y   0 ) / S a Normal (0,1)
• Note that our chosen significance levels are only approximate.
As n gets large, the tn‐1 distribution converges to the standard
normal distribution. • When the sample size is large, the actual significance level will be very close to 5%.

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
Example C.5: Race C. Asymptotic C. Asymptotic

Discrimination in Hiring
D. P-Value Computing and Using p‐Values D. P-Value
E. CI & HT E. CI & HT

• =B‐W is the difference in the probability that blacks and whites receive job • The traditional requirement of choosing the significance level ahead of

offers. time means that different researchers could wind up with different
 is the population mean of the variable Y=B‐W where B and W are binary variables. conclusions,
– although they use the same set of data and same procedures.
• Testing
H0: =0
• p‐value of the test
H1: 0

• Given n=241, y  .133se( y )  .48 / 241  .031 • p‐value of the test

It is the largest significance level at which we fail to reject the null hypothesis.
• The t statistic for testing H0: =0
t = ‐.133/.031 = ‐4.29
• p‐value of the test
It is the smallest significance level at which we reject the null hypothesis.
• Critical value = ‐2.58 (two‐sided test; =.01)
• t<‐2.58 There is very strong evidence against H0 in favor of H1.

• One sided Test: Let H0: =0 in a Normal(,2). • Interpretation: t=1.52 and p‐value=.065
The test statistic is T  n  Y / S

• The observed value of T for our sample is
– The largest significance level at which we carry out the
t = 1.52
test and fail to reject the H0 is .065.
– The probability that we observe a value of T as large as
• The p‐value is the area to the right 1.52 when the null hypothesis is true.
hand side of 1.52, which is
– If we carry out the test at the significance level above
.065, we reject the null hypothesis.
p‐value = P(T>1.52H0)
= 1 – (1.52) = .0655
– The smallest significance level at which we reject the null
hypothesis is .065.
where () is the standard normal – We would observe the value of T as large as 1.52 due to
cumulative distribution function (cdf). chance 6.5% of the time.

A. Fundamentals A. Fundamentals
B. HT N(0,1) Example C.6: Effect of Training Grants on B. HT N(0,1)
C. Asymptotic C. Asymptotic
Computing and Using p‐Values D. P-Value Worker Productivity D. P-Value
E. CI & HT E. CI & HT
(one tail test)
•  is the average change in scrap rates; n=20. Note that the change in scrap rates has a
normal distribution.
• Interpretation: t=2.85 and (n is large)
• Hypothesis
p‐value= 1 – (2.85) =.0022 H0: = 0 (Training grants have no effect)
H1:  <0
– If the null hypothesis is true, we observe a value of T as large as 2.85 n=20 (for one‐tail test)
with probability .002.
– If we carry out the test at the significance level above .002, we • If we carry out the test at
reject the null hypothesis. the significance level
– The smallest significance level at which we reject the null above 0.023, we reject the
hypothesis is .002. null hypothesis.

• The smallest significance
level at which we reject
the null hypothesis is
0.023.

Two sided alternative • Given
• For t testing about population means, the p‐value is t value of the test statistic n=241; = ‐.133; se( ) = .48/ = .031
P(Tn‐1>t) = 2P(Tn‐1>t) Tn-1 t random variable
• The t statistic for testing H0: =0
t = ‐.133/.031 = ‐4.29
• P‐value is computed by finding the area to the right oftand multiplying the For nonnormal distribution,
area by two. the exact p‐value can be
• Hull Hypothesis and Two‐sided alternative • If Z is a standard normal random variable difficult to obtain, but we
H0: =0 against P(Z<‐4.29)  0 can find asympototic p‐
H1:   0 values by using the same
• Hull Hypothesis and Two‐sided alternative calculations.
• If we carry out the test at the significance P-value H0: =0 against
= 0.023+0.023
level above 0.046, we reject the null = 0.046 H1:   0
hypothesis.
Area = p-value = 023
• The smallest significance level at which we
reject the null hypothesis is 0.046. • There is very strong evidence against H0 in favor of H1.
2.13 – Note that Critical value = ‐2.58 (=.01)

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
C. Asymptotic The Relationship between Confidence C. Asymptotic
Computing and Using p‐Values D. P-Value
E. CI & HT Interval and Hypothesis Testing
D. P-Value
E. CI & HT

• Rejection rules for t value: Summary • The confidence interval and hypothesis testing are linked.
1) For H1:  > 0, the rejection rule is t>c and the p‐value is P(T>t). • Assume =.05. The confidence interval can be used to test two sided
alternatives. Suppose
2) For H1:  < 0, the rejection rule is t<‐c and the p‐value is P(T<t). H0:  = 0
3) For H1:  ≠ 0, the rejec on rule is t>c and the p‐value is P(T>t). H1:   0

• Rejection Rule
– If 0 does not in the confidence interval, we reject the null hypothesis at the
5% level.
• Rejection rules for p‐value: Summary
– If the hypothesized value of , 0, lies in the confidence interval, we fail to
Choose significance level,  reject the null hypothesis at the 5% level.
1) We reject the H0 at the 100% level if
p‐value <  • After a confidence interval is constructed, many values of 0 can be
2) We fail to reject H0 at the 100% level if tested.
p‐value   – Since a confidence interval contains more than one value, there are many
null hypotheses that can be rejected.
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
107 108
VI. Hypothesis Testing Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat VI. Hypothesis Testing Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
Example C.8:  Training Grants C. Asymptotic Practical Significance and C. Asymptotic
D. P-Value D. P-Value
and Worker Productivity E. CI & HT Statistical Significance E. CI & HT

• A 95% confidence interval for the mean change in scrap rate  is • Three covered evidences of population parameters include
[‐2.28, ‐0.02] 1) point estimates,
• Since zero is excluded from this interval, at 5% level, we reject 2) confidence intervals and
H0: =0 against H1:   0
3) hypothesis tests.
• If H0: =‐2, we fail to reject the null hypothesis.
• In empirical analysis, we should also put emphasis on the magnitudes of
the point estimates!
• Don’t say
We “accept” the null hypothesis H0: =‐1.0 at the 5% significance level.
• Statistical significance depends on the size of the test statistic, not
• This is because in the same set of data, there are usually many hypotheses on the size of .
that cannot be rejected.
– It depends on the ratio of  to its standard error: t  y / se( y )
• For example, • The test statistic could be large because se( ) is large or is large.
– it is logically incorrect to say that H0: =‐1 and H0: = ‐2 are both “accepted.”
– It is possible that neither are rejected.
– Thus, we say “fail to reject”.

A. Fundamentals A. Fundamentals

Practical Significance and B. HT N(0,1)
C. Asymptotic
Example C.9: Effect of Free Way B. HT N(0,1)
C. Asymptotic
D. P-Value Width on Commute Time D. P-Value
Statistical Significance E. CI & HT E. CI & HT

• Let Y denote the change in commute time, measured in minutes, for
commuters before and after a freeway was widened.
• Note that the magnitude and sign of the test statistic • Assume YNormal(, 2)
determine the statistical significance. • Hypotheses:
 H0: =0
 H1: <0
• Practical significance depends on the magnitude of .
– The estimate can be statistically significant without being large, • Given n=900, y  3.6; sample sd  32.7 se( y )  32.7 / 900  1.09
especially when we work with large sample sizes. • The t statistic for testing H0: =0
t = ‐3.6/1.09 = ‐3.30
p‐value=.005

• Statistical Significance: We conclude that the estimated reduction in comunte time

had a statistically significant effect on average commute time.
• Practical Significance: The estimated reduction in average commute time is only 3.6
minutes.
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
VI. Hypothesis Testing Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 111 VI. Hypothesis Testing Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 112
VII. Remarks on Notation Remarks on Notation
• In the main text, we use a simpler convention that is widely used in
econometrics.
• We have been careful to use standard conventions to denote
random variable W • If  is a population parameter, the notation will be used to denote
estimator W both an estimator and an estimate of 
estimate w
• = theta hat

• Distinguishing between an estimator and estimate (outcome Example:
of the random variable W) is important for understanding • If the population parameter is , then
various concepts in denotes an estimator or estimate of .
– Estimation and
– Hypothesis Testing.
• If the parameter is 2, then
denotes an estimator or estimate of 2.
I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks I. Random S. II. Finite S. III. Asymptotic S. IV. Parameter E. V. Interval E. & Confidence I. VI. Hypothesis T VII. Remarks
VII. Remarks on Notation Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 113 VII. Remarks on Notation Fundamentals of Mathematical Statistics . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat 114

A. Fundamentals A. Fundamentals
B. HT N(0,1) B. HT N(0,1)
Problem C.6 C. Asymptotic
D. P-Value Problem C.6 continue
C. Asymptotic
D. P-Value
E. CI & HT E. CI & HT

C.6 You are hired by the governor to study whether a tax on
liquor has decreased average liquor consumption in your (iii) Now, suppose your sample size is n = 900 and you obtain the
estimates = ‐ 32.8 and s = 466.4.
state. You are able to obtain, for a sample of individuals
Calculate the t statistic for testing H0 against H1; obtain the p‐value for
selected at random, the difference in liquor consumption (in the test. (Because of the large sample size, just use the standard
ounces) for the years before and after the tax. For person i normal distribution tabulated in Table G.1.) Do you reject H0 at the 5%
who is sampled randomly from the population, Yi denotes the level? at the 1% level? [ans.]
change in liquor consumption. Treat these as a random
sample from a Normal(, 2) distribution. (iv) Would you say that the estimated fall in consumption is large in
magnitude? Comment on the practical versus statistical significance of
this estimate. [ans.]
(i) The null hypothesis is that there was no change in average liquor
consumption. State this formally in terms of . [ans.]
(ii) The alternative is that there was a decline in liquor consumption; state the (v) What has been implicitly assumed in your analysis about other
alternative in terms of . [ans.] determinants of liquor consumption over the two‐year period in order
to infer causality from the tax change to liquor consumption? [ans.]

VI. Hypothesis Testing VII. Remarks on Notation VI. Hypothesis Testing VII. Remarks on Notation
115 116
Fundamentals of Mathematical Statistics: Part Three . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat Fundamentals of Mathematical Statistics: Part Three . Intensive Course in Mathematics and Statistics . Chairat Aemkulwat
y

Problem C.6 (i) (ii) Problem C.6 (iii)
(iii)
Yi – the change in liquor consumption. • The standard error of  is
se( ) =  s/ = 466.4/30 = 15.55.
Yi are a random sample from a Normal (, 2)
• Therefore, the t statistic for testing H0: µ = 0 is
t =   /se( ) = – 32.8/15.55
= –2.11.
(i)   • We obtain the p‐value as
P(Z  –2.11),
– H0: µ = 0. where Z ~ Normal(0,1).

(ii) • These probabilities are in Table G.1:
p‐value = .0174.
– H1: µ < 0. • (=.05) Because the p‐value is below .05, we reject H0 against the one‐sided
alternative at the 5% level.
• (=.01) We do not reject at the 1% level because p‐value = .0174 > .01.

Problem C.6 (iv)
(iv)
• The estimated reduction, about 33 ounces, does not seem large for an
entire year’s consumption.
– If the alcohol is beer, 33 ounces is less than three 12‐ounce cans of beer.
– Even if this is hard liquor, the reduction seems small.
– (On the other hand, when aggregated across the entire population, alcohol
distributors might not think the effect is so small.)
(v)
• The implicit assumption is that other factors that affect liquor
consumption – such as income, or changes in price due to
transportation costs, are constant over the two years.

Problem C.7 C. Rule of Thumb

D. Asymptotic CI Problem C.7 continue … D. P-Value
E. CI & HT

C.7 The new management at a bakery obs Wb Wa D=Wa-Wb (ii) Formally state the null hypothesis that there has been no

claims that workers are now more
productive than they were under old
1 8.3 9.25 0.95
change in average wages. In particular, what is E(Di) under H0?
management, which is why wages have
2 9.4 9 -0.4
If you are hired to examine the validity of the new
“generally increased.” Let Wib be Worker
3

4
9

10.5
9.25

10
0.25

-0.5
management’s claim, what is the relevant alternative
i’s wage under the old management and hypothesis in terms of E(Di)? [ans.]
let Wia be Worker i’s wage after the 5 11.4 12 0.6

change. The difference is 6 8.75 9.5 0.75

Di = Wia ‐ Wib.
7

8
10

9.5
10.25

9.5
0.25

0
• (iii) Test the null hypothesis from part (ii) against the stated
Assume that the Di are a random sample 9 10.8 11.5 0.7
alternative at the 5% and 1% levels. [ans.]
from a Normal(, 2) distribution. 10 12.55 13.1 0.55

(i) Using the following data on 15 workers,
11

12
12

8.65
11.5

9
-0.5

0.35
• (iv) Obtain the p‐value for the test in part (iii). [ans.]
construct an exact 95% confidence 13 7.75 7.75 0

interval for . [ans.] 14 11.25 11.5 0.25

15 12.65 13 0.35

mean 10.16667 10.40667 0.24

Problem C.7 (i) Problem C.7 (i)
Date: 05/07/07 Time: 07:57
(i)
Sample: 1 15
The average increase in wage is ̅ = .24, or 24
obs Wb Wa D=Wa-Wb
•
cents. 1 8.3 9.25 0.95 WB WA D
2 9.4 9 -0.4 Mean 10.16667 10.40667 0.24
• The sample standard deviation is about s = 3 9 9.25 0.25
Median 10 10 0.25
.451, n = 15, se( ̅ ) = .1164. 4 10.5 10 -0.5
Maximum 12.65 13.1 0.95
5 11.4 12 0.6
Minimum 7.75 7.75 -0.5
• From Table G.2, the 97.5th percentile in the t14 6 8.75 9.5 0.75

distribution is 2.145. Std. Dev. 1.569084 1.595291 0.450872

7 10 10.25 0.25
Skewness 0.175376 0.290842 -0.34947
8 9.5 9.5 0

• So the 95% CI is [d  c  S / n , d  c  S / n ] 9 10.8 11.5 0.7

Kurtosis 1.810807 2.022774 2.161199

10 12.55 13.1 0.55

= .24  2.145(.1164), 11 12 11.5 -0.5

Jarque-Bera 0.960754 0.80833 0.745067
= or about –.010 to .490. 12 8.65 9 0.35 Probability 0.61855 0.667534 0.688986
(ii) 13 7.75 7.75 0

• If µ = E(di) then 14 11.25 11.5 0.25 Sum 152.5 156.1 3.6

H0: µ = 0. 15 12.65 13 0.35 Sum Sq. Dev. 34.46833 35.62933 2.846
• The alternative is that management’s claim is mean 10.16667 10.40667 0.24
true: H1: µ > 0.
Observations 15 15 15

Problem C.7 (iv)
Hypothesis Testing for DI
Date: 05/07/07 Time: 08:03
Sample: 1 15
Good Luck!
Included observations: 15
Test of Hypothesis: Mean = 0.000000

Sample Mean = 0.240000

Sample Std. Dev. = 0.450872
FT19, PT15: See you around!
Method Value Probability
t-statistic 2.061595 0.0583

View / Test of Descriptive Stats / Simple Hypothesis Tests

Sta 341 Class Notes Final
No ratings yet
Sta 341 Class Notes Final
120 pages
Slides SM 1
No ratings yet
Slides SM 1
51 pages
Statistical Inference Frequentist
No ratings yet
Statistical Inference Frequentist
25 pages
Notes
No ratings yet
Notes
140 pages
FECO Note 1 - Review of Statistics: Xuan Chinh Mai
No ratings yet
FECO Note 1 - Review of Statistics: Xuan Chinh Mai
13 pages
Statistics 1B Lecture Notes: Author: T. Farrar
No ratings yet
Statistics 1B Lecture Notes: Author: T. Farrar
129 pages
Review of Probability and Statistics
No ratings yet
Review of Probability and Statistics
34 pages
POL 571: Convergence of Random Variables: 1 Random Sample and Statistics
No ratings yet
POL 571: Convergence of Random Variables: 1 Random Sample and Statistics
7 pages
Transition To MATH503
No ratings yet
Transition To MATH503
12 pages
Point Estimation: Statistics (MAST20005) & Elements of Statistics (MAST90058) Semester 2, 2018
No ratings yet
Point Estimation: Statistics (MAST20005) & Elements of Statistics (MAST90058) Semester 2, 2018
12 pages
Lecture 4
No ratings yet
Lecture 4
66 pages
Theme5 Sampling Distribution
No ratings yet
Theme5 Sampling Distribution
29 pages
Lecture1 Review&Intro
No ratings yet
Lecture1 Review&Intro
34 pages
MIR - Ivchenko G. I., Medvedev Yu. and Chistyakov A. - Problems in Mathematical Statistics - 1991
100% (4)
MIR - Ivchenko G. I., Medvedev Yu. and Chistyakov A. - Problems in Mathematical Statistics - 1991
282 pages
Part 2-1 Random Samples Sampling Distributions - Notes
No ratings yet
Part 2-1 Random Samples Sampling Distributions - Notes
8 pages
Chapter 5 Introduction To Statistical Inference
No ratings yet
Chapter 5 Introduction To Statistical Inference
9 pages
Lec # 3
No ratings yet
Lec # 3
47 pages
ParameterEstimation Slides
No ratings yet
ParameterEstimation Slides
40 pages
Sampling & Sampling Distributions
No ratings yet
Sampling & Sampling Distributions
26 pages
ECON4150 - Introductory Econometrics Lecture 2: Review of Statistics
No ratings yet
ECON4150 - Introductory Econometrics Lecture 2: Review of Statistics
41 pages
Chap 1 Sampling Distributions
No ratings yet
Chap 1 Sampling Distributions
14 pages
Basic Statistic
No ratings yet
Basic Statistic
20 pages
STA 303 Lec 1
No ratings yet
STA 303 Lec 1
5 pages
Week 7
No ratings yet
Week 7
113 pages
Lecture 5 - Part II
No ratings yet
Lecture 5 - Part II
9 pages
Random Sampling, Statistics, and Estimators
No ratings yet
Random Sampling, Statistics, and Estimators
9 pages
Estimation of The Population Mean
No ratings yet
Estimation of The Population Mean
16 pages
Chapter 1 - Estimation Theory
No ratings yet
Chapter 1 - Estimation Theory
166 pages
Topic06 Written
No ratings yet
Topic06 Written
15 pages
MIT2 854F10 Stats
No ratings yet
MIT2 854F10 Stats
38 pages
X, ..., X, X, ..., X X, X, ..., X: Basic Statistics
No ratings yet
X, ..., X, X, ..., X X, X, ..., X: Basic Statistics
29 pages
Lecture 5
No ratings yet
Lecture 5
39 pages
Sampling Notes Part 01
No ratings yet
Sampling Notes Part 01
13 pages
Lecture 3: Sampling and Sample Distribution
No ratings yet
Lecture 3: Sampling and Sample Distribution
30 pages
Statistical Inferences: Dr. Olivia Carrillo Gamboa
No ratings yet
Statistical Inferences: Dr. Olivia Carrillo Gamboa
16 pages
Bizstat ssn2
No ratings yet
Bizstat ssn2
55 pages
ECON 327 Lecture 03 Statistics
No ratings yet
ECON 327 Lecture 03 Statistics
95 pages
04.sampling Distributions of The Estimators
No ratings yet
04.sampling Distributions of The Estimators
32 pages
Chapter 1 - Estimation Theory - 1-731
No ratings yet
Chapter 1 - Estimation Theory - 1-731
731 pages
Chapter 1 - Estimation Theory
No ratings yet
Chapter 1 - Estimation Theory
1,440 pages
Book Down
No ratings yet
Book Down
17 pages
Module02 Slides Print 1
No ratings yet
Module02 Slides Print 1
65 pages
Math Statistics
No ratings yet
Math Statistics
4 pages
DPBS 1203 Business and Economic Statistics
No ratings yet
DPBS 1203 Business and Economic Statistics
21 pages
Sampling Distribution
No ratings yet
Sampling Distribution
41 pages
L06 Inference
No ratings yet
L06 Inference
48 pages
Topic 5: Probability Bounds and The Distribution of Sample Statistics
No ratings yet
Topic 5: Probability Bounds and The Distribution of Sample Statistics
16 pages
Lecture 4 - Part III
No ratings yet
Lecture 4 - Part III
49 pages
P Sample
No ratings yet
P Sample
3 pages
Session 31 - Sample Statistics
No ratings yet
Session 31 - Sample Statistics
28 pages
Course: Statistical Inference & Applications: Instructor in Charge
No ratings yet
Course: Statistical Inference & Applications: Instructor in Charge
30 pages
Statistical Inference
No ratings yet
Statistical Inference
82 pages
Chap5 Statistical Inference
No ratings yet
Chap5 Statistical Inference
38 pages
Estimation: M. Shafiqur Rahman
No ratings yet
Estimation: M. Shafiqur Rahman
31 pages
MATH 403 Engineering Data Analysis 95 132
No ratings yet
MATH 403 Engineering Data Analysis 95 132
38 pages
Chapter10 Solutions
No ratings yet
Chapter10 Solutions
62 pages
Session2 QTII 24
No ratings yet
Session2 QTII 24
31 pages
USP-NF 228 Ethylene Oxide and Dioxane
No ratings yet
USP-NF 228 Ethylene Oxide and Dioxane
3 pages
Summative Test Reading and Writing
No ratings yet
Summative Test Reading and Writing
3 pages
Science 7 q3 Balanced and Unbalanced Forces Week 1
No ratings yet
Science 7 q3 Balanced and Unbalanced Forces Week 1
10 pages
Economics
No ratings yet
Economics
322 pages
Plastic Waste Management
No ratings yet
Plastic Waste Management
9 pages
Team5 Final
No ratings yet
Team5 Final
24 pages
23MIPrelimQP (H2 Chem Paper 3)
No ratings yet
23MIPrelimQP (H2 Chem Paper 3)
26 pages
Ocb & LMX
No ratings yet
Ocb & LMX
12 pages
Leadership Theories: MBA 645 Leadership in Organizations Jeff Shay University of Montana
No ratings yet
Leadership Theories: MBA 645 Leadership in Organizations Jeff Shay University of Montana
16 pages
Activity 2 - Formula Writing and Nomenclature of Inorganic Compounds
No ratings yet
Activity 2 - Formula Writing and Nomenclature of Inorganic Compounds
2 pages
Division Memorandum No. 0555, S. 2024 - Reiteration On The Implementation of Modular Distance Learning As Provided in DepEd Order No. 037, S. 2022.
No ratings yet
Division Memorandum No. 0555, S. 2024 - Reiteration On The Implementation of Modular Distance Learning As Provided in DepEd Order No. 037, S. 2022.
2 pages
Blue - Print - Pre-Board - 2 - Pol. - SC - XII
No ratings yet
Blue - Print - Pre-Board - 2 - Pol. - SC - XII
4 pages
To Technical Paper III
No ratings yet
To Technical Paper III
10 pages
Nas5311 Spec
No ratings yet
Nas5311 Spec
4 pages
SOP For Chiller Unit ON - OFF
No ratings yet
SOP For Chiller Unit ON - OFF
5 pages
Li 2008
No ratings yet
Li 2008
4 pages
Amity University, Mumbai Aibas: Title: Learned Optimism Scale
No ratings yet
Amity University, Mumbai Aibas: Title: Learned Optimism Scale
9 pages
GREENER - Energy Efficiency in Buildings
No ratings yet
GREENER - Energy Efficiency in Buildings
4 pages
Sentence Structure
No ratings yet
Sentence Structure
3 pages
FDT Crusherun L 1
No ratings yet
FDT Crusherun L 1
1 page
Mayo Clinic Internal Medicine Board Review 10th
No ratings yet
Mayo Clinic Internal Medicine Board Review 10th
303 pages
Flexural Behaviour of RC One-Way Slabs Reinforced Using PAN Based Carbon Textile Grid
No ratings yet
Flexural Behaviour of RC One-Way Slabs Reinforced Using PAN Based Carbon Textile Grid
13 pages
1316020AL TM Primary 3A Unit 1 Numbers To 100,000
No ratings yet
1316020AL TM Primary 3A Unit 1 Numbers To 100,000
23 pages
Thermographic Scanning Fact Sheet
No ratings yet
Thermographic Scanning Fact Sheet
3 pages
EAPP A Sample Critique Paper
No ratings yet
EAPP A Sample Critique Paper
4 pages
Important Questions MARKETING MANAGEMEMT
No ratings yet
Important Questions MARKETING MANAGEMEMT
3 pages
ULG II Master Thesis Traussnig PDF
No ratings yet
ULG II Master Thesis Traussnig PDF
90 pages
MSDS LDPE - LLDPE Version 7 EN
No ratings yet
MSDS LDPE - LLDPE Version 7 EN
11 pages
Art of Agile Dev
No ratings yet
Art of Agile Dev
45 pages
Leweke Et Al 2016 Dynamics and Instabilities of Vortex Pairs
No ratings yet
Leweke Et Al 2016 Dynamics and Instabilities of Vortex Pairs
37 pages

Appendix C Mathematical Statistics 2015

Uploaded by

Appendix C Mathematical Statistics 2015

Uploaded by

Outline:

• Parameters are constants that determine the directions and Example: Results:

• An estimator W of a parameter  can be expressed as

• Unbiasedness ensures that the probability distribution of an estimator has a • Summary:

Suppose W1 and W2 both are unbiased This implies that the probability that W1 is

Relative Efficiency A. Unbiasedness A. Unbiasedness

• The estimator is efficient relative to Y1 for estimating .

C.1 Let Y1, Y2, Y3, and Y4 be independent, identically distributed (ii) Now, consider a different estimator of :

(i) What are the expected value and variance of in terms of  (iii) Based on your answers to parts (i) and (ii), which

Consistency B. Asy. Normality Consistency B. Asy. Normality

1. The distribution Wn becomes 3. When Wn is consistent, we say that  is

P(Wn‐ ) >   0 as n  

– Thus, Var( n)  0 as n ‐> 

Consistency B. Asy. Normality

• As n  , and X* are consistent estimators of .

Example: Central Limit Theorem

where (z) is the standard normal cumulative distribution function (cdf) The variable Zn is the standardized

has an asymptotic standard normal distribution. “a” stands for “asymptotically”

The exact distributions of W1 = [(n‐1)/n] and

(i) Show that W1 and W2 are both biased estimators of  and find

• Because plim(W1) = µ and plim(W2) = µ/2, W1 is • Var(W2) = Var( )/4 = 2/(4n).

Example: Population mean 1)This is a consistent estimator of XY.

• Given the sample average is an unbiased and consistent estimator of , Example: Sample covariance

Example: Population correlation • Maximum likelihood Estimator (MLE) 1. In the discrete case, this is

2) The log of the product is the sum

Summary Unbiased Consistent Efficiency Example:

• Interval estimation comes in when we make statement involving • The standardized version of has a standard normal distribution.

• Interval estimation uses information on the point estimate and the standard • The random interval is P(Y 1.96 / n <   Y  1.96 / n )  0.95

–  Normal(, 2 /n). • (Y  1.96 / n , Y  1.96 / n )

• s/ is the point estimate of sd( or the standard error of . • A 95% confidence interval for the mean

• For some applications, the population is nonnormal. • Matched pairs analysis – each person in a pair interviewed for the same job.

• Unbiased estimators of B and W are and the fractions of interviews for which

1 8.3 9.25 0.95

change. The difference is 6 8.75 9.5 0.75

• Find an approximate 95% confidence interval for  7 10 10.25 0.25

• A 99% CI for  = B ‐ w is ‐.133  2.58(.482/(241)½

interval for . [ans.] 14 11.25 11.5 0.25

mean 10.16667 10.40667 0.24

2 9.4 9 -0.4 Mean 10.16667 10.40667 0.24

se( ̅ ) = .1164. 11 12 11.5 -0.5

12 8.65 9 0.35 Probability 0.61855 0.667534 0.688986

14 11.25 11.5 0.25 Sum 152.5 156.1 3.6

• A significance level is the probability of committing Type I error. () = P(Reject H0) = 1 – P(Type II)

• The test statistic T is some function of the random sample. • To test hypothesis about the mean  from a

Testing Hypotheses about the B. HT N(0,1)

• Given n=241, y   .133 se ( y )  .48 / 241  . 031 • The rejection rule is

• =B‐W is the difference in the probability that blacks and whites receive job • The traditional requirement of choosing the significance level ahead of

• Given n=241, y  .133se( y )  .48 / 241  .031 • p‐value of the test

• Statistical Significance: We conclude that the estimated reduction in comunte time

Problem C.7 C. Rule of Thumb

C.7 The new management at a bakery obs Wb Wa D=Wa-Wb (ii) Formally state the null hypothesis that there has been no

change. The difference is 6 8.75 9.5 0.75

interval for . [ans.] 14 11.25 11.5 0.25

mean 10.16667 10.40667 0.24

distribution is 2.145. Std. Dev. 1.569084 1.595291 0.450872

• So the 95% CI is [d  c  S / n , d  c  S / n ] 9 10.8 11.5 0.7

10 12.55 13.1 0.55

= .24  2.145(.1164), 11 12 11.5 -0.5

• If µ = E(di) then 14 11.25 11.5 0.25 Sum 152.5 156.1 3.6

Sample Mean = 0.240000

View / Test of Descriptive Stats / Simple Hypothesis Tests

You might also like