Scribd
Upload
55 views13 pages

Lecture 03 & 04 EC5002 CLT & Normal Distribution 2015-16

This document provides an overview of key concepts in statistics including the central limit theorem, confidence intervals, and hypothesis testing. Specifically: - The central limit theorem states that the sampling distribution of the sample mean approaches the normal distribution as the sample size increases, regardless of the population distribution. - Confidence intervals use sample information and the normal distribution to calculate a range that is likely to contain the true population parameter with a given level of confidence, such as 95%. - Hypothesis testing involves formulating a null hypothesis, calculating a test statistic, determining critical values, and either rejecting or failing to reject the null based on where the test statistic falls relative to the critical values. The goal is to minimize type 1

Uploaded by

nishit0157623637
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
55 views13 pages

Lecture 03 & 04 EC5002 CLT & Normal Distribution 2015-16

This document provides an overview of key concepts in statistics including the central limit theorem, confidence intervals, and hypothesis testing. Specifically: - The central limit theorem states that the sampling distribution of the sample mean approaches the normal distribution as the sample size increases, regardless of the population distribution. - Confidence intervals use sample information and the normal distribution to calculate a range that is likely to contain the true population parameter with a given level of confidence, such as 95%. - Hypothesis testing involves formulating a null hypothesis, calculating a test statistic, determining critical values, and either rejecting or failing to reject the null based on where the test statistic falls relative to the critical values. The goal is to minimize type 1

Uploaded by

nishit0157623637
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

Economics Quantitatively Treated 2 (EQT 2)

(EC5002)
Lecture 3 and 4
The Central Limit Theorem, the normal distribution,
confidence intervals and hypothesis testing

Ref.: Gujarati (2009), Appendix D

JUS:EC5002:2013-14:L3

Learning outcomes:
By the end of this class and completion of the
reading you should be able to know:
The Central Limit Theorem
The sample distribution of X-bar
Confidence Intervals
Hypothesis testing

JUS:EC5002:2013-14:L3

The Central Limit Theorem (CLT)


X - a normally distributed random variable, X ~ N( , 2x )
x
With expected value (mean):

and variance:

2x

Sample this distribution: sample size n


2

The mean value of X: X ~ N( , x )


x n
CLT: Given ANY prob. distribution of a RV, mean x and var. 2x
whether it is normally distributed or not, the sampling distribution
of the sample mean approaches the normal distribution, as long as
the sample size is large.
2

That is: X ~ N( , x )
x n
JUS:EC5002:2013-14:L3

X-
2
Z~
where

Standardizing the sample mean: Z X - X - ~ N(0,1)


2
n
n

Standard Normal:

Example: X~N(20, 4); ~N(20, 4/n): = 20 mpg, (2/n) = (4/n),


what is the probability of a random sample of size 25 cars will give
a mean value of (a) greater than 21 mpg and (b) less than 18 mpg.
X
X 20 X 20
We know Z
~ N(0,1), so Z

0.4
n
2 25
21 - 20
a) P( X 21) P(Z
) P(Z 2.5) 0.0062 (very unlikely)
0.4
b) P( X 18) P(Z

JUS:EC5002:2013-14:L3

18 - 20
) P(Z 5) 0 (so unlikely to be virtually zero)
0.4

Confidence intervals
Use sample information and find a confidence interval which is
very likely to contain the true population parameter X.
Statisticians often choose a probability of 0.95 (or 95%), i.e. 1 in
20 chance of being wrong.
The sample mean has a normal distribution with a large enough
sample size, use standardised normal (Z) distribution.
From Z table: Z values of -1.96 to + 1.96 include 95% of all Z
values, i.e. P(-1.96 < Z < + 1.96) = 0.95

JUS:EC5002:2013-14:L3

Confidence intervals
X X
P(-1.96
1.96) 0.95
n

X X X X
note that

P(-1.96

(X ) n
X

1.96) 0.95

Multiply by
P(-1.96 ( X ) n 1.96 ) 0.95 divide by n
X

divide by n

P -1.96
X X 1.96
0.95
n
n

JUS:EC5002:2013-14:L3

Substract X

X

X
-1.96

1
.
96

X
n
n

0.95

Multiply by (-1)

P X 1.96

n


P X 1.96

n


JUS:EC5002:2013-14:L3

X 1.96

n


X 1.96

n

0.95

0.95

Example : X 19.5, n 25, 4, population mean is unknown


2

P X -1.96
X X 1.96

n
n


P( 19.5-1.96 0.4)
P( 18.716
JUS:EC5002:2013-14:L3

0.95

19.5 1.96 0.4) 0.95

20.284) 0.95
8

When population mean or variance is unknown, use sample


information
2
( X i X )
2
2
Sample variance : s
, so s s
n 1
s is called standard error (rather than the standard deviation. It is
standard deviation of a population, but standard error of a
sample). Now given that we only know the sample variance
which we substitute into
X
X
the expression Z
~ N (0,1) to get :
~ t ( n 1)
n
s n
Example : X 11.5, n 50, s 3.0456, (Gujarati) , df 49
P( 2.0096 t 2.0096) 0.95
3.0456
3.0456
P( 11.5 2.0096(
) 11.5 2.0096)(
) 0.95
X
50
50
P( 10.63 12.36) 0.95
X

JUS:EC5002:2013-14:L3

Hypothesis Testing
Hypothesize, X = 13, i.e. test whether sample information we
have leads us to suppose X = 13. This is called
NULL of maintained hypothesis: H0: X = 13
The alternative hypothesis, HA (or H1): X 13 (two tailed test)
or
HA: X > 13 or (< 13) (one tailed test).
If X were long way from 13 then we would not maintain
H0: X = 13. Use standardise distance (to take out scaling factors)
by dividing the s.e. This gives us a t value, that is the distance
expressed in numbers of s.e. from 0. This is known as,
X
X
Test statistics
~ t with degrees of freedom (d.f.) n - 1
s n
JUS:EC5002:2013-14:L3

10

Example (see above) : X 11.5, n 50, s 3.0456, d.f. 49


11.5 13
3.4826
Null: H0: X = 13; Alternative: HA 13, t
3.0456 50
The critical values (within which we accept null hypothesis) are
the t values for 49 d.f. and a probability, or 0.95, namely -2.0096
and +2.0096
If the test statistic is large sample mean is long way from the
hypothesized population mean the null is unlikely to be true.

The test statistic lies in the tails (rejection region) reject the
null hypothesis.
In our example, test statistic is -3.4826, inside the rejection
reject the null that true population mean is 13.
The degree of certainty: 0.99 (99%) with about 50 d.f.: CV=-2.68
to + 2.68
One tail test
JUS:EC5002:2013-14:L3

11

Error
The degree of certainty: 0.95 (95%) chances of being wrong:
0.05 (5%). This is known as the level of significance or size of the
test. We choose this level so we are choosing the chance of being
wrong. The ways of being wrong are:
TYPE I: rejecting the null (H0) given that it is true.
TYPE II: accepting the null (H0) given that is false.
When we choose the size of a test we are choosing the probability
of committing a type I error.
In general, when testing we want to minimize the probability of
committing type II error given the chosen size of the test.

JUS:EC5002:2013-14:L3

12

Step by step approach to hypothesis testing


Formulate an hypothesis, H0 and HA
Establish the critical values using the t distribution (make sure you
know the degrees of freedom) and choose a level of significance.
Establish a decision rule, e.g. if the test statistic is within the
acceptance region: accept H0. If the test statistic is outside the
acceptance region then reject H0.
Calculate the test statistic from the sample information:
t

X
s

Make a decision.
JUS:EC5002:2013-14:L3

13

You might also like