University of Gondar

1
College of medicine and health science
Department of Epidemiology and Biostatistics

 Estimation and Hypothesis Testing

Prepared By: Department of Epidemiology and Biostatistics

 December, 2018

University of Gondar, Ethiopia

 Objectives
2

 After complete this session you will be able to do

 Parameter estimations
 Pointestimate
 Confidence interval
 Hypothesis testing
 Z-test
 T-test

 Analysis of variance
 Introduction # 1
3

 Inferential is the process of generalizing or drawing

conclusions about the target population on the basis
of results obtained from a sample.
 Introduction #2
4

 Before beginning statistical analyses

 it is essential to examine the distribution of the variable for
skewness (tails),
 kurtosis (peaked or flat distribution), spread (range of the
values) and
 outliers (data values separated from the rest of the data).
 Information about each of these characteristics
determines to choose the statistical analyses and can
be accurately explained and interpreted.
 Sampling Distribution
5

 The frequency distribution of all these samples forms the

sampling distribution of the sample statistic
 Sampling distribution .......
6
 Three characteristics about sampling distribution of a statistic
 its mean
 its variance
 its shape

 Due to random variation different samples from the same

population will have different sample means.
 If we repeatedly take sample of the same size n from a
population the means of the samples form a sampling
distribution of means of size n is equal to population mean.
 In practice we do not take repeated samples from a population
i.e. we do not encounter sampling distribution empirically, but it
is necessary to know their properties in order to draw statistical
inferences.
 The Central Limit Theorem
7

 Regardless of the shape of the frequency distribution of a

characteristic in the parent population,
 the means of a large number of samples (independent

observations) from the population will follow a normal

distribution (with the mean of means approaches the population
mean μ, and standard deviation of σ/√n ).
 Inferential statistical techniques have various assumptions that
must be met before valid conclusions can be obtained
 Samples must be randomly selected.
 sample size must be greater (n>=30)
 the population must be normally or approximately normally
distributed if the sample size is less than 30.
 Sampling Distribution......
8

  
 Sampling Distribution ..........
9

  
Standard deviation and Standard error

Standard deviation is a measure of variability between

individual observations (descriptive index relevant to
mean)
Standard error refers to the variability of summary
statistics (e.g. the variability of the sample mean or a
sample proportion)
Standard error is a measure of uncertainty in a sample
statistics i.e. precision of the estimate of the estimator
 Parameter Estimations
11

 In parameter estimation, we generally assume that the

underlying (unknown) distribution of the variable of interest is
adequately described by one or more (unknown) parameters,
referred as population parameters.

 As it is usually not possible to make measurements on every

individual in a population, parameters cannot usually be
determined exactly.

 Instead we estimate parameters by calculating the

corresponding characteristics from a random sample estimates
.
 the process of estimating the value of a parameter from
information obtained from a sample.
 Estimation

Estimation is a procedure in which we use the information

included in a sample to get inferences about the true parameter
of interest.
 An estimator is a sample statistic that used to estimate the
population parameter while an estimate is the possible values
that a given estimator can assume.
Properties of a good estimator
Sample statistic Corresponding population
parameter
(Sample mean) μ (population mean)
S2 (sample variance) σ2 (population variance)
S (sample Standard deviation) σ (population standard deviation)
(Sample proportion) P (Population proportion)
A desirable property of a good estimator is the following
 It should be unbiased: The expected value of the estimator must be
equal to the parameter to be estimated.
 It should be consistent: as the sample size increase, the value of the
estimator should approaches to the value of the parameter estimated.
 It should be efficient: the variance of the estimator is the smallest.
 It should be sufficient: the sample from which the estimator is
 Types of Estimation
14

There are two types of estimation:

1. Point estimation: It uses the information in the sample to

arrive at a single number (that is called an estimate) that
is intended to be close to the true value of the parameter.

2. Interval estimation: It uses the information of the

sample to end up at an interval (i.e. construct 2 endpoints)
that is intended to enclose the true value of the parameter.
 Point Estimation
15

  

 x
p =
n
Example

  
 Some BLUE estimators
17
 Interval Estimation
18

 However the value of the sample statistic will vary from

sample to sample therefore, to simply obtain an
estimate of the single value of the parameter is not
generally acceptable.
 We need also a measure of how precise our estimate is likely to
be.
 We need to take into account the sample to sample variation of
the statistic.

 A confidence interval defines an interval within which

the true population parameter is like to fall (interval
estimate).
 Confidence Intervals…
19

 Confidence interval therefore takes into account the sample to sample

variation of the statistic and gives the measure of precision.
 The general formula used to calculate a Confidence interval is Estimate
± K × Standard Error, k is called reliability coefficient.
 Confidence intervals express the inherent uncertainty in any medical
study by expressing upper and lower bounds for anticipated true
underlying population parameter.
 The confidence level is the probability that the interval estimate will
contain the parameter, assuming that a large number of samples are
selected and that the estimation process on the same parameter is
repeated.
 Confidence intervals…
20

 Most commonly the 95% confidence intervals are calculated,

however 90% and 99% confidence intervals are sometimes
used.
 The probability that the interval contains the true population
parameter is (1-α)100%.
 If we were to select 100 random samples from the population
and calculate confidence intervals for each, approximately 95
of them would include the true population mean B (and 5
would not)
 Confidence interval ……
21

A (1-α) 100% confidence interval for unknown population

mean and population proportion is given as follows;

 
[x  z . , x  z . ]
2 n 2 n
 
[ p  z . p(1 p) / n, p  z . p(1 p) / n]
2 2
 Interval estimation
22
23
24
25
 Confidence intervals…
26

 The 95% confidence interval is calculated in such a way that, under the
conditions assumed for underlying distribution, the interval will contain true
population parameter 95% of the time.
 Loosely speaking, you might interpret a 95% confidence interval as one which
you are 95% confident contains the true parameter.
 90% CI is narrower than 95% CI since we are only 90% certain that the interval
includes the population parameter.
 On the other hand 99% CI will be wider than 95% CI; the extra width meaning
that we can be more certain that the interval will contain the population
parameter. But to obtain a higher confidence from the same sample, we must
be willing to accept a larger margin of error (a wider interval).
 Confidence intervals…
27

 For a given confidence level (i.e. 90%, 95%, 99%) the

width of the confidence interval depends on the
standard error of the estimate which in turn depends on
the
 1. Sample size:-The larger the sample size, the narrower the

confidence interval (this is to mean the sample statistic will

approach the population parameter) and the more precise our
estimate. Lack of precision means that in repeated sampling
the values of the sample statistic are spread out or scattered.
The result of sampling is not repeatable.
 Confidence intervals…
28

  
Confidence interval for a single mean

CI =

Most commonly, we used to compute 95% confidence

interval, however, it is possible to compute 90% and 99%
confidence interval estimation.
 Table 1: Normal distribution
Area between 0 and z

30   0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.0 0.0000 0.0040 0.0080 0.0120 0.0160 0.0199 0.0239 0.0279 0.0319 0.0359
0.1 0.0398 0.0438 0.0478 0.0517 0.0557 0.0596 0.0636 0.0675 0.0714 0.0753
0.2 0.0793 0.0832 0.0871 0.0910 0.0948 0.0987 0.1026 0.1064 0.1103 0.1141
0.3 0.1179 0.1217 0.1255 0.1293 0.1331 0.1368 0.1406 0.1443 0.1480 0.1517
0.4 0.1554 0.1591 0.1628 0.1664 0.1700 0.1736 0.1772 0.1808 0.1844 0.1879
0.5 0.1915 0.1950 0.1985 0.2019 0.2054 0.2088 0.2123 0.2157 0.2190 0.2224
0.6 0.2257 0.2291 0.2324 0.2357 0.2389 0.2422 0.2454 0.2486 0.2517 0.2549
0.7 0.2580 0.2611 0.2642 0.2673 0.2704 0.2734 0.2764 0.2794 0.2823 0.2852
0.8 0.2881 0.2910 0.2939 0.2967 0.2995 0.3023 0.3051 0.3078 0.3106 0.3133
0.9 0.3159 0.3186 0.3212 0.3238 0.3264 0.3289 0.3315 0.3340 0.3365 0.3389
1.0 0.3413 0.3438 0.3461 0.3485 0.3508 0.3531 0.3554 0.3577 0.3599 0.3621
1.1 0.3643 0.3665 0.3686 0.3708 0.3729 0.3749 0.3770 0.3790 0.3810 0.3830
1.2 0.3849 0.3869 0.3888 0.3907 0.3925 0.3944 0.3962 0.3980 0.3997 0.4015
1.3 0.4032 0.4049 0.4066 0.4082 0.4099 0.4115 0.4131 0.4147 0.4162 0.4177
1.4 0.4192 0.4207 0.4222 0.4236 0.4251 0.4265 0.4279 0.4292 0.4306 0.4319
1.5 0.4332 0.4345 0.4357 0.4370 0.4382 0.4394 0.4406 0.4418 0.4429 0.4441
1.6 0.4452 0.4463 0.4474 0.4484 0.4495 0.4505 0.4515 0.4525 0.4535 0.4545
1.7 0.4554 0.4564 0.4573 0.4582 0.4591 0.4599 0.4608 0.4616 0.4625 0.4633
1.8 0.4641 0.4649 0.4656 0.4664 0.4671 0.4678 0.4686 0.4693 0.4699 0.4706
1.9 0.4713 0.4719 0.4726 0.4732 0.4738 0.4744 0.4750 0.4756 0.4761 0.4767
2.0 0.4772 0.4778 0.4783 0.4788 0.4793 0.4798 0.4803 0.4808 0.4812 0.4817
2.1 0.4821 0.4826 0.4830 0.4834 0.4838 0.4842 0.4846 0.4850 0.4854 0.4857
2.2 0.4861 0.4864 0.4868 0.4871 0.4875 0.4878 0.4881 0.4884 0.4887 0.4890
2.3 0.4893 0.4896 0.4898 0.4901 0.4904 0.4906 0.4909 0.4911 0.4913 0.4916
 Confidence interval ……
31

 If the population standard deviation is unknown and the

sample size is small (<30), the formula for the
confidence interval for sample mean is:

 x is the sample mean

 s is the sample standard deviation
 n is the sample size
 t is the value from the t-distribution with (n-1) degrees of
freedom
 The t Distribution
32
df t0.100 t0.050 t0.025 t0.010 t0.005
--- ----- ----- ------ ------ ------
1 3.078 6.314 12.706 31.821 63.657 t D is trib utio n: d f = 1 0
2 1.886 2.920 4.303 6.965 9.925
3 1.638 2.353 3.182 4.541 5.841 0 .4
4 1.533 2.132 2.776 3.747 4.604
5 1.476 2.015 2.571 3.365 4.032
6 1.440 1.943 2.447 3.143 3.707 0 .3
7 1.415 1.895 2.365 2.998 3.499 Area = 0.10 Area = 0.10
8 1.397 1.860 2.306 2.896 3.355

f(t)

}
0 .2
9 1.383 1.833 2.262 2.821 3.250
10 1.372 1.812 2.228 2.764 3.169
11 1.363 1.796 2.201 2.718 3.106 0 .1
12 1.356 1.782 2.179 2.681 3.055
13 1.350 1.771 2.160 2.650 3.012
14 1.345 1.761 2.145 2.624 2.977 0 .0
15 1.341 1.753 2.131 2.602 2.947 -1.372 0 1.372
-2.228 2.228
16 1.337 1.746 2.120 2.583 2.921

}
}
17 1.333 1.740 2.110 2.567 2.898 t
18 1.330 1.734 2.101 2.552 2.878
19 1.328 1.729 2.093 2.539 2.861 Area = 0.025 Area = 0.025
20 1.325 1.725 2.086 2.528 2.845
21 1.323 1.721 2.080 2.518 2.831
22
23
1.321
1.319
1.717
1.714
2.074
2.069
2.508
2.500
2.819
2.807 Wheneverisisnot
Whenever notknown
known(and
(andthe
thepopulation
populationisis
24
25
1.318
1.316
1.711
1.708
2.064
2.060
2.492
2.485
2.797
2.787 assumednormal),
assumed normal),thethecorrect
correctdistribution
distributiontotouse
useisis
26
27
1.315
1.314
1.706
1.703
2.056
2.052
2.479
2.473
2.779
2.771 thet tdistribution
the distributionwith
withn-1
n-1degrees
degreesofoffreedom.
freedom.
28
29
1.313
1.311
1.701
1.699
2.048
2.045
2.467
2.462
2.763
2.756 Note,however,
Note, however,that
thatforforlarge
largedegrees
degreesofoffreedom,
freedom,
30
40
1.310
1.303
1.697
1.684
2.042
2.021
2.457
2.423
2.750
2.704 thet tdistribution
the distributionisisapproximated
approximatedwell
wellbybythe
theZZ
60
120
1.296
1.289
1.671
1.658
2.000
1.980
2.390
2.358
2.660
2.617 distribution.
distribution.
 1.282 1.645 1.960 2.326 2.576
Point and Interval Estimation of the Population Proportion (p)

We will now consider the method for estimating the binomial

proportion p of successes, that is, the proportion of elements in a
population that have a certain characteristic.
A logical candidate for a point estimate of the population
x
proportion p is the sample proportion ˆ
p 
n, where x is the number
of observations in a sample of size n that have the characteristic
of interest. As we have seen in sampling distribution of
proportions, the sample proportion is the best point estimate of
the population proportion.
 Proportion…
34

 The shape is approximately normal provided n is sufficiently

large - in this case, nP > 5 and nQ > 5 are the requirements
for sufficiently large n ( central limit theorem for proportions) .
 The point estimate for population proportion π is given by þ.

 A (1-α)100% confidence interval estimate for the unknown

population proportion π is given by:

CI =  p  Z  (1   ) / n , p  Z

 (1   ) / n 
   
 2 2 
 If the sample size is small, i.e. np < 5 and nq < 5, and the
population standard deviations for proportion are not given,
then the confidence interval estimation will take t-distribution
instead of z as:
 Example 1:
35

 A SRS of 16 apparently healthy subjects yielded the following values of urine

excreted (milligram per day);
0.007, 0.03, 0.025, 0.008, 0.03, 0.038, 0.007, 0.005, 0.032, 0.04, 0.009, 0.014,
0.011, 0.022, 0.009, 0.008
Compute point estimate of the population mean
If x 1 , x 2 , ..., x n are n observed values , then
n

 xi
0 . 295
x = i =1
  0 . 01844
n 16

Construct 90%, 95%, 98% confidence interval for the mean

(0.01844-1.65x0.0123/4, 0.01844+1.65x0.0123/4)=(0.0134, 0.0235)
(0.01844-1.96x0.0123/4, 0.01844+1.96x0.0123/4)=(0.0124, 0.0245)
(0.01844-2.33x0.0123/4, 0.01844+2.33x0.0123/4)=(0.0113, 0.0256 )
 Example 2

The mean diastolic blood pressure for 225 randomly

selected individuals is 75 mmHg with a standard
deviation of 12.0 mmHg. Construct a 95% confidence
interval for the mean
Solution
n=225
mean =75mmhg
Standard deviation=12 mmHg
confidence level 95%
The 95% confidence interval for the unknown population mean is
given
95%CI = (75 ±1.96x12/15) = (73.432,76.56)
 Example 2:
37

AAstock
stockmarket
marketanalyst
analystwants
wantstotoestimate
estimatethetheaverage
averagereturn
returnon
onaa
certainstock.
certain stock. AArandom
randomsample
sampleof of15
15days
daysyields
yieldsan
anaverage
average
(annualized)return
(annualized) returnof
of x  10.37 andandaastandard
standarddeviation
deviationof
ofss==
3.5.Assuming
3.5. Assumingaanormal
normalpopulation
populationof ofreturns,
returns,give
giveaa95%
95%confidence
confidence
intervalfor
interval forthe
theaverage
averagereturn
returnononthis
thisstock.
stock.

df
---
t0.100
-----
t0.050
-----
t0.025
------
t0.010
------
t0.005
------ The critical value of t for df = (n -1) = (15 -1)
1
.
3.078
.
6.314
.
12.706
.
31.821
.
63.657
. =14 and a right-tail area of 0.025 is:
t 0.025  2.145
. . . . . .
. . . . . .
13 1.350 1.771 2.160 2.650 3.012
14 1.345 1.761 2.145 2.624 2.977 The corresponding confidence interval or
15 1.341 1.753 2.131 2.602 2.947 s
. . . . . .
interval estimate is: x  t 0 . 025
.
.
.
.
.
.
.
.
.
.
.
. n
35
.
 10.37  2.145
15
 10.37  1.94
  8.43,12.31
 Example 3:
38

 In a survey of 300 automobile drivers in one city, 123 reported that

they wear seat belts regularly. Estimate the seat belt rate of the city
and 95% confidence interval for true population proportion.
 Answer : p= 123/300 =0.41=41%
n=300,
Estimate of the seat belt of the city at 95%
CI = p ± z ×(√p(1-p) /n) =(0.35,0.47)
 Example 4:
In a sample of 400 people who were questioned regarding their participation in sports,
160 said that they did participate. Construct a 98 % confidence interval for P, the
proportion of P in the population who participate in sports.
Solution:
Let X= be the number of people who are interested to participate in sports.
X=160, n=400, =0.02, Hence
 Z 2  Z 0.01  2.33

ˆ X 160 P (1  P ) 0 .4 ( 0 .6 )
P   0.4  P2ˆ    0.0245
n 400 n 400

As a result, an approximate 98% confidence interval for P is given by:

Pˆ (1  Pˆ ) Pˆ (1  Pˆ )   ( 0 .4  ( 2 .33 * 0 .0245 )), ( 0 .4  ( 2 . 33 * 0 .0245 
 Pˆ  Z 2  P  Pˆ  Z 2 )
n n   0 . 345 ,0 .457 

Hence, we can conclude that about 98% confident that the true proportion of people in
the population who participate in sports between 34.5% and 45.7%.
 HYPOTHESIS TESTING
40

Introduction
 Researchers are interested in answering many types of

questions. For example, A physician might want to

know whether a new medication will lower a person’s
blood pressure.

 These types of questions can be addressed through

statistical hypothesis testing, which is a decision-
making process for evaluating claims about a
population.
 Hypothesis Testing
41

 The formal process of hypothesis testing provides us with a

means of answering research questions.

 Hypothesis is a testable statement that describes the nature of the

proposed relationship between two or more variables of interest.

 In hypothesis testing, the researcher must defined the population

under study, state the particular hypotheses that will be
investigated, give the significance level, select a sample from the
population, collect the data, perform the calculations required for
the statistical test, and reach a conclusion.
 Idea of hypothesis testing
42
 type of Hypotheses
43
 Null hypothesis (represented by HO) is the statement about the value of the
population parameter. That is the null hypothesis postulates that ‘there is no
difference between factor and outcome’ or ‘there is no an intervention effect’.
 Alternative hypothesis (represented by HA) states the ‘opposing’ view that
‘there is a difference between factor and outcome’ or ‘there is an intervention
effect’.
 Methods of hypothesis testing
44

 Hypotheses concerning about parameters which may or may

not be true
 Examples
• The mean GPA of this class is 3.5!
• The mean height of the Gondar College of Medical
Sciences (GCMS) students is 1.63m.
• There is no difference between the distribution of Pf and Pv
malaria in Ethiopia (are distributed in equal proportions.)
 Steps in hypothesis testing
45

1 2

Identify the null hypothesis H0 and Choose a. The value should be small, usually less
than 10%. It is important to consider the
the alternate hypothesis HA. consequences of both types of errors.

3
Select the test statistic and determine 4
its value from the sample data. This
value is called the observed value of Compare the observed value of the statistic to the
the test statistic. Remember that t critical value obtained for the chosen a.
statistic is usually appropriate for a
small number of samples; for larger
number of samples, a z statistic can 5
work well if data are normally Make a decision.
distributed. 6
Conclusion
 Test Statistics
46
 Because of random variation, even an unbiased sample may not
accurately represent the population as a whole.
 As a result, it is possible that any observed differences or
associations may have occurred by chance.
 A test statistics is a value we can compare with known distribution
of what we expect when the null hypothesis is true.
 The general formula of the test statistics is:

Observed _ Hypothesized
Test statistics = value value .

Standard error
 The known distributions are Normal distribution, student’s distribution ,
Chi-square distribution ….
 Critical value
47

 The critical value separates the critical region from the noncritical region
for a given level of significance
 Decision making
48

 Accept or Reject the null hypothesis

 There are 2 types of errors

Type of decision H0 true H0 false

Reject H0 Correct decision (1-

Type I error (a)
β)

Accept H0 Correct decision (1-a) Type II error (β)

 Type I error is more serious error and it is the level of significant

 power is the probability of rejecting false null hypothesis and it is
given by 1-β
49
50
51
 Types of testes

One tailed test a Critical

H0: m = m0
Value(s)
H1: m <
0
m0 Rejection Regions
a
H0: m = m0
H1: m > m0 0
a/2
H0: m = m0
H1: m ¹
0
m0 Two tailed test
 Hypothesis testing about a Population mean
(μ)
53

Two Tailed Test:

The large sample (n > = 30) test of hypothesis about a population mean μ is
as follows
1 H 0 :    0 (   0 )
H A :  1   0 (   0 )
x  0
z cal 

n
z tabulated  z  for two tailed test
2

 if | z cal | z tab reject H o

Decision : 
 if | z cal | z tab do not reject H o
 Steps in hypothesis testing…..
54

If the test statistic does not fall in the

If the test statistic falls in the critical
critical region:
region:
Conclude that there is not enough
Reject H0 in favour of HA.
evidence to reject H0.
 One tailed tests
55

2 H 0 :    0 (   0 )
H A :  1   0 (   0 )
x  
z cal  0
, z tabulated  z for one tailed test

n
 if z cal   z tab reject H o
Decision : 
 if z cal   z tab do not reject H o

3 H 0 :    0 (   0 )
H A :  1   0 (   0 )
 if z cal  z tab reject H o
Decision : 
 if z cal  z tab do not reject H o
 The P- Value
56

 In most applications, the outcome of performing a hypothesis test

is to produce a p-value.
 P-value is the probability that the observed difference is due to
chance.
 A large p-value implies that the probability of the value observed,
occurring just by chance is low, when the null hypothesis is true.

 That is, a small p-value suggests that there might be sufficient

evidence for rejecting the null hypothesis.
 The p value is defined as the probability of observing the
computed significance test value or a larger one, if the H0
hypothesis is true. For example, P[ Z >=Zcal/H0 true].
 P-value……
57

 A p-value is the probability of getting the

observed difference, or one more extreme, in the
sample purely by chance from a population where
the true difference is zero.

 If the p-value is greater than 0.05 then, by

convention, we conclude that the observed
difference could have occurred by chance and there
is no statistically significant evidence (at the 5%
level) for a difference between the groups in the
population.
How to calculate P-value
o Use statistical software like SPSS, SAS……..
o Hand calculations

—obtained the test statistics (Z Calculated or t-

calculated)
—find the probability of test statistics from standard
normal table
—subtract the probability from 0.5
—the result is P-value
Note if the test two tailed multiply 2 the result.
 P-value and confidence interval
59

 Confidence intervals and p-values are based upon the same

theory and mathematics and will lead to the same conclusion
about whether a population difference exists.

 Confidence intervals are referable because they give

information about the size of any difference in the population,
and they also (very usefully) indicate the amount of uncertainty
remaining about the size of the difference.

 When the null hypothesis is rejected in a hypothesis-testing

situation, the confidence interval for the mean using the same
level of significance will not contain the hypothesized mean.
 The P- Value …..
60

 But for what values of p-value should we reject the null

hypothesis?
 By convention, a p-value of 0.05 or smaller is considered
sufficient evidence for rejecting the null hypothesis.
 By using p-value of 0.05, we are allowing a 5% chance of
wrongly rejecting the null hypothesis when it is in fact
true.

 When the p-value is less than to 0.05, we often say that the
result is statistically significant.
 Hypothesis testing for single population mean
61

EXAMPLE 5: A researcher claims that the mean of the IQ for 16

students is 110 and the expected value for all population is 100 with
standard deviation of 10. Test the hypothesis .
 Solution
1. Ho:µ=100 VS HA:µ≠100
2. Assume α=0.05
3. Test statistics: z=(110-100)4/10=4
4. z-critical at 0.025 is equal to 1.96.
5. Decision: reject the null hypothesis since 4 ≥ 1.96
6. Conclusion: the mean of the IQ for all population is different
from 100 at 5% level of significance.
 Example 6:
62

Suppose that we have a population mean 3.1 and n=20

people x  4.5 and s  5 . 5 found and , our test statistic is
1. Ho:  3.1
HA:   3.1
2. α = 0.5 at 95% CI
x   4.5  3.1
3. t  s  5.5 1.14 t0.05,19  2.09
n 20

4. the observed value of the test statistic falls with in the

range of the critical values
5. we accept Ho and conclude that there is no enough
 Cont….
63

A 95% confidence interval for the mean is

x  t0.05,19s / n  4.5  2.09(5.5 / 20)  (1.93,7.07)

Note that this interval includes the hypothesis

value of 3.1
 Hypothesis testing for single proportions
64

Example 7: In the study of childhood abuse in psychiatry patients, brown

found that 166 in a sample of 947 patients reported histories of physical or sexual
abuse.
a) constructs 95% confidence interval
b) test the hypothesis that the true population proportion is 30%?
 Solution (a)
 The 95% CI for P is given by
p (1  p )
p  z 
2 n
0 . 175  0 . 825
 0 . 175  1 . 96 
947
 0 . 175  1 . 96  0 . 0124
 [ 0 . 151 ; 0 .2 ]
 Example……
65


To the hypothesis we need to follow the steps
Step 1: State the hypothesis
Ho: P=Po=0.3
Ha: P≠Po ≠0.3
Step 2: Fix the level of significant (α=0.05)
Step 3: Compute the calculated and tabulated value of the test statistic

p  Po 0 .175  0 .3  0 .125
z cal      8.39
p (1  p ) 0 .3( 0 .7 ) 0 .0149
n 947
z tab  1 .96
 Example……
66

 Step 4: Comparison of the calculated and tabulated values of the

test statistic
 Since the tabulated value is smaller than the calculated value of the
test the we reject the null hypothesis.
 Step 6: Conclusion
 Hence we concluded that the proportion of childhood abuse in
psychiatry patients is different from 0.3

 If the sample size is small (if np<5 and n(1-p)<5) then use student’s
t- statistic for the tabulated value of the test statistic.
67

Statistical Inference Based on

Two Samples
Statistical Inferences Based on Two Samples

Comparing Two Population Means;

 Independent Samples: Variances Known
 Independent Samples: Variances Unknown

Paired Difference Experiments

 Paired/matched/repeated sampling

Comparing Two Population Proportions

 Large, Independent Samples case
Comparing Two Population Means;
Case-1: Independent Samples, Variances Known

  
Comparing Two Population Means;
Independent Samples, Vars Known cont’d…

  

 12  22
 x1  x 2  
n1 n2
Comparing Two Population Means;
Ind’t Samples, Vars Known cont’d…

 A (1 – ) 100% confidence interval for the difference in

populations µ1–µ2 is;
 2 
1  2
2
 x1  x2   z 2  
 n1 n2 

 In testing hypothesis, the z value can then be calculated as;

z 
 x1  x 2  D 0

 12  2
2

n1 n 2 ; where Do = (µ1 – µ2)o
 Hypothesis testing for two sample means
72

 The steps to test the hypothesis for difference of means is the same
with the single mean
Step 1: state the hypothesis
Ho: µ1-µ2 =0
VS
HA: µ1-µ2 ≠0, HA: µ1-µ2 <0, HA: µ1-µ2 >0
Step 2: Significance level (α)
Step 3: Test statistic
(x  y)  (1   2 )
z cal 
 12  22

n1 n2
 Hypothesis …
73

z tabulated  z  for two tailed test

2

z tabulated  z  for one tailed test

 if | z cal | z tab reject H o
For H A :  1   2  0 
 if | z cal | z tab do not reject H o
 if z cal   z tab reject H o
For H A :  1   2  0 
 if z cal  z cal do not reject H o
 if z cal  z cal reject H o
For H A :  1   2  0 
 if z cal  z cal do not reject H o
 Example 8:
74

 A researchers wish to know if the data they have collected provide

sufficient evidence to indicate a difference in mean serum uric acid
levels between normal individual and individual with down’s
syndrome. The data consists of serum uric acid readings on 12
individuals with down’s syndrome and 15 normal individuals. The
means are 4.5mg/100ml and 3.4 mg/100ml with standard deviation
of 2.9 and 3.5 mg/100ml respectively.

H O : 1   2  0
H A : 1   2  0
 SOLUTION
75

(x  y)  (1   2 ) ( 4 .3  3 .4 )  0
z cal  
 2
 2
2 .9 2 3 .5 2
1
 2

n1 n2 12 15
1 .6 1 .6
   5 . 33
1 . 5178 1 . 23
z   z 0 . 025  1 . 96
2
Comparing Two Population Means;
case-2: Independent Samples, Variances Unknown

  
 Comparing Two Population Means;
Ind’t Samples, Vars Unknown cont’d…

A. Assume that the unknown variances; σ12 = σ22 = σ2

 The pooled estimate of σ2 is the weighted average of the two
sample variances, s12 and s22
 The pooled estimate of σ2 is denoted by sp 2 .

s 2p 
 n1  1 s12   n 2  1 s 22
n1  n 2  2

 The estimate of the population standard deviation of the

sampling distribution is;
2 1 1 
 x1  x2  s p   
 n1 n2 
 Comparing Two Population Means;
Ind’t Samples, Vars Unknown cont’d…

 The sampling distribution of is in this case

approximately normal when both n1 and n2 are large (>30),
irrespective of the distribution of the population. So, in this
case; the CI and Z-cal are given as;
 A (1 – ) 100% CI for µ1 – µ2 is;
  1 1 
  x1  x 2   Z  2 s 2
p   
  n1 n2  


 The calculated value of z will be;

z 
 x1  x 2  D 0

 1 1 
s 2
p    where Do = (µ1 – µ2)o
 n1 n2 
 Comparing Two Population Means;
Ind’t Samples, Vars Unknown cont’d…

  

  1 1 
  x1  x 2   t 2 ( n1  n 2  2 ) s 
2
p   
  n1 n2  


t 
 x1  x 2  D 0

 1 1 
s 2
p   
 n1 n2 
Comparing Two Population Means;
Ind’t Samples, Vars Unknown cont’d…

  

z 
x1  x 2   D 0

s 12 s 2
 2
n 1 n 2

 x1  x  D df 
s 2
1 /n 1  s /n 2
2
2  2

s   
t  2 0
2 2
2 2
s 12

s 22 /n 1 s /n 2
n1 n2
1
 2

n1  1 n2  1
 Comparing Two Population Means;
Case-3: Paired/matched/repeated sampling

 Rises from two different processes on same study units (e.g.

"before” and “after” treatments) or two different processes on
paired/matched study units ( e.g. Pair matched case control
studies).
 Use of the same/matched individuals, eliminates any
differences in the individuals themselves (confounding factors).
 Inference concerning the difference between two population
means is similar to one population mean; except that we will be
manipulating on the dis here.
Paired sampling cont’d…

  
Paired sampling cont’d…

 If the population of differences is normally distributed with

mean d
 A (1- )100% confidence interval for
µd = µ1 - µ2 is:
 sd 
 d  t  /2 
 n

 Where for a sample of size n, t/2 is based on n – 1 degrees of

freedom.

 but Z-test can be used if the sample size is large

(n1=n2=n>30).
Paired sampling cont’d…

  
Paired sampling cont’d…
  
Hypothesis testing for two proportions

 Suppose that n1 and n2 are large enough so that;

 n ·p ≥5, n ·(1 - p )≥5, n ·p ≥5, and n ·(1 – p )≥5
1 1 1 1 2 2 2 2

 Then the population of all possible values of p̂1 - p̂2;

 Has approximately a normal distribution
 Has mean µp̂1 - p̂2 = p1 – p2
p1 1  p1  p 2 1  p 2 
 p̂1  p̂ 2  
 Has standard deviation; n1 n2
Hypothesis testing for two proportions

 A (1 – ) 100% confidence interval for p1 - p2;

 p̂1 1  p̂1  p̂ 2 1  p̂ 2  
  p̂1  p̂ 2   z  2  
 n1 n2 

 The test statistic is;

z=
 p̂1  p̂ 2   D0
 p̂1  p̂2
where Do = (P1-P2)0
 Hypothesis testing for two proportions
88

 To test the hypothesis

Ho: π1-π2 =0
VS
HA: π1-π2 ≠0
The test statistic is given by

( p 1  p 2 )  ( 1   2 )
z cal 
p 1 (1  p 1 ) p 2 (1  p 2 )

n1 n2
 Small sample size
89

  
 Comparing Two Population Proportions cont’d…

Example 10: A study was conducted to look at the effects of oral

contraceptives (OC) on heart disease in women 40–44 years of age. It is
found that among n1 = 500 current OC users, 13 develop a myocardial
infarction (MI) over a three-year period, while among n2 = 1000 non-OC
users, seven develop a MI over a three-year period. Then;

A. Construct a 95% confidence interval for the difference of MI rates

between OC-users and non-users.

B. Can you conclude that rate of MI is significantly greater among OC

users? (Report the P-value for your test)
 Comparing Two Population Proportions cont’d…

 Solution: The estimation (CI) for the difference of

population proportions should be formed using the
following formula (for a 95% confidence interval):

A.

Where ≈ 0.005.

 pˆ1 1  pˆ1  pˆ 2 1  pˆ 2  
 pˆ1  pˆ 2   z 2    0.019  1.96* 0.0035
 n1 n2 
 The 95% CI for the difference is = (0.012, 0.026)
92

ANOVA
 Introduction
 Here in the case of two independent sample t-test, we
have one continuous dependent variable (interval/ratio
data) and;

 one nominal or ordinal independent variable with only

two categories

 In this last case (i.e. two independent

sample t-test), what if there are more
than two categories for the
independent variable we have?
 Are the birth weights of children in different geographical
regions the same?

 Are the responses of patients to different medications and

placebo different?

 Are people with different age groups have different

proportion of body fat?

 Do people from different ethnicity have the same BMI?

One way-Analysis Of Variance

 All the above research questions have one common

characteristic: That is each of them has two variables: one
categorical and one quantitative

 Main question: Are the averages of the quantitative variable

across the groups (categories) the same?

 Because there is only one categorical independent variable

which has two or more categories (groups), the name one
way ANOVA comes.
 One-way ANOVA cont…
96

 Also called Completely Randomized Design

 Experimental units (subjects) are assigned randomly

to treatments/groups. Here subjects are assumed to
be homogeneous
 Analysis of variance cont…
97
 One way ANOVA is a method for testing the hypothesis:
There is no difference between two or more population means
(usually at least three); or there is no difference between a number of
treatments

 More formally, we can state hypotheses as:

H0: There is no difference among the mean of treatments effects

HA: There is difference at least between two treatments effects
or
Ho: µ1 = µ2 = µ3 =…. = µa (if there are ‘a’ groups)
HA: at least one group mean is different
Why Not Just Use t-tests?
 Since t-test considers two groups at a time, it will be tedious when many
groups are present

 Conducting multiple t-tests can lead also to severe inflation of the Type I
error rate (false positives) and is not recommended

 However, ANOVA is used to test for differences among several means

without increasing the Type I error rate

 The ANOVA uses data from all groups at a time to estimate standard
errors, which can increase the power of the analysis
Assumptions of One Way ANOVA)

 The data are normally distributed or the samples have come

from normally distributed populations and are independent.

 The variance is the same in each group to be compared (equal

variance).

 Moderate departures from normality may be safely ignored, but

the effect of unequal standard deviations may be serious.

 In the latter case, transforming the data may be useful.

Analysis of variance cont…
 We test the equality of means among groups by using the
variance

 The difference between variation within groups and

variation between groups may help us to compare the
means

 If both are equal, it is likely that the observed difference

is due to chance and not real difference
Note that:
Total Variability = Variability between + Variability
within
 Analysis of variance
 Basic model: Data are deviations from the
μ global mean, μ:(The Linear Model)
Xij = μ + Ɛij
 Sum of vertical deviations squared is the total
sum of squares = SSt
G-1 G-2

A2

One way model: Data are
deviations from treatment means,
Ais:

A1
Xij = μ + Ai + Ɛij
 Sum of vertical deviations squared
= SSe
G-1 G-2
 Decomposing the total variability
102
n a n a n a
 Total SS = Σ Σ (xij – )2 = ΣiΣjxij2 - (ΣiΣjxij)2 /na = SST
i=1 j=1

n a n a a n
 Within SS = Σ Σ (xij – j
)2 = ΣiΣjxij2 - Σj(Σixij)2/n = SSW
i=1 j=1

n a a n
 Between SS = Σ Σ ( i j
– )2 = Σj(Σixij)2/n - (ΣiΣjxij)2 /na = SSB
i=1 j=1

This is assuming each of the ‘a’ groups has equal size, ‘n’.

SST = SSW + SSB

Data of one way ANOVA
Groups/variable
G-1 G-2 G-3 ….. G-a
X11 X12 X13 ….. X1a
Participant
X21 X22 X23 ….. X2a
X31 X32 X33 ….. X3a
. . . . .
. . . . .
. . . . .
s

Xn1 Xn2 Xn3 …. Xna

Totals T.1 T.2 T.3 …. T.a
Computational formula
T= ΣiΣjxij2 Correction Factor = CF = (ΣiΣjxij)2 /na = T2../na
A = Σj(Σixij)2/n = Σj(T.j)2/n if the groups’ (cells’) size are equal, or
A = Σj(Σixij)2/nj = Σj(T.j)2/nj ; if unequal group size

Where, Xij = ith observation in the jth group of the table

i = 1, 2, 3,…, nj, j = 1, 2, 3,…,a, Σjnj = N
 Sum of squares and ANOVA Table

Source of df SS MS F
variation
Between groups a-1 SSB = A - CF SSB/(a-1) MSB/MSW
Within groups na-a SSW = T - A SSW/(na –a)
Total na-1 SST = T - CF

 If there are real differences among groups’ means, the between

groups variation will be larger than the within variation
Example on one-way ANOVA
The following table shows the red cell folate levels (μg/l) in three groups of
cardiac bypass patients who were given three different levels of nitrous oxide
ventilation. (Level of nitrous oxide for group I > group II’s > group III’s)

Group I Group II Group III

(n=8) (n=9) (n=5)
243 206 241
251 210 258
275 226 270
291 249 293
347 255 328
354 273
380 285
392 295
309
Total=2533 2308 1390
Mean =316.6 256.4 278.0
SD = 58.7 37.1 33.8
Example Cont….
We can see the box plot just to have some
impression about it
 Example cont…
Ho: μ1 = μ2 = μ3
HA: Differences exist between at least two of the means
Source of variation df SS Mean F P
square

Between groups 2 15516 7758

3.71 0.044

Within groups 19 39716 2090

Total 21 55232

Since the P-value is less than 0.05, the null hypothesis is rejected
 Pair-wise comparisons of group means post hoc tests or multiple comparisons

 ANOVA test tells us only whether there is statistically significant

difference among groups means, but

 It doesn’t tell us which particular groups are significantly

different

 To identify them, we use either a priori (pre-planed) or post hoc

tests
 Pair-wise comparisons of group means (post hoc tests) cont…

 Whether to use a priori or post hoc tests depends on whether the

researcher has previously stated the hypotheses to test.

 If you have honestly stated beforehand the comparisons between

individual pairs of means which you intend to make, then you are
entitled to use a priori test such as a t-test.
 In this case, only one pair of groups or few will be tested

 However, when you look at the data it may seem worth

comparing all possible pairs. In this case, a post hoc test such as
Scheffe, Benferroni (modified t-test), Tuckey methods, Least
Squares Difference (LSD), etc will be employed.
Benferroni method or Modified t-test (Steps)

I. Find tcalc for the pairs of groups of interest (to be compared)

II. The modified t-test is based on the pooled estimate of

variance from all the groups (which is the residual variance
in the ANOVA table), not just from pair being considered.

III. If we perform k paired comparisons, then we should

multiply the P value obtained from each test by k; that is, we
calculate P' = kP with the restriction that P' cannot exceed 1.
Where, , that is the number of possible comparisons
Benferroni method or Modified t-test
 Returning to the red cell folate data given above, the residual
standard deviation is = 45.72.

(a) Comparing groups I and II

t = (316.6 - 256.4) / (45.72 x √(1/9 +1/8)
= 2.71 on 19 degrees of freedom.

 The corresponding P-value = 0.014 and the

corrected P value is P' = 0.014x3
= 0.042
Group I and II are different
 Benferroni method or Modified t-test
 
(b) Comparing groups I and III

 t = (316.6 - 278.0) / (45.72 x √(1/8+1/5)

= 38.6/26.06
= 1.48 on 19 degrees of freedom.

 The corresponding P value = 0.1625 and

 The corrected P value is P' = 0.1625x3

= 0.4875
Group I and III are not different
Benferroni method or Modified t-test

(c) Comparing Groups II and III

 t = (278 - 256.4) / (45.72 x √(1/5+1/9)

= 21.6/25.5
= 0.85 on 19 degrees of freedom.

 The corresponding P value = 0.425 and the corrected P value is P' =

1.00

Group I and III are not different

Therefore, the main explanation for the difference between the groups
that was identified in the ANOVA is thus the difference between groups
I and II.
 Which post hoc method Shall I use?
 The post hoc tests differ from one another in how they calculate
the p value for the mean difference between groups.

 Least Squares Difference (LSD) is the most liberal of the post hoc
tests and has a high Type I error rate. It is simply multiple t-tests

 The Scheffé test uses the F distribution rather than the t

distribution of the LSD tests and is considered more conservative.

It has a high Type II error rate but

is considered appropriate when
there are a large number of groups
to be compared.
Which post hoc method Shall I use? cont…

 The Bonferroni approach uses a series of t tests ( that is

the LSD technique) but corrects the significance level
for multiple testing by dividing the significance levels by
the number of tests being performed

 Since this test corrects for the number of comparisons

being performed, it is generally used when the number
of groups to be compared is small.
Which post hoc method Shall I use? Cont..

 Tukey’s Honesty Significance Difference (Tukey’s HSD)

test also corrects for multiple comparisons, but it considers
the power of the study to detect differences between groups
rather than just the number of tests being carried out;

 That is, it takes into account sample size as well as the

number of tests being performed.

 This makes it preferable when there are a large number of

groups being compared, since it reduces the chances of a
Type I error occurring.
 ANOVA - a recapitulation.

 ANOVA is a parametric test, examining whether the

means differ between 2 or more populations.

 It generates a test statistic F, which can be thought of

as a signal: noise ratio. Thus large values of F indicate
a high degree of pattern within the data and imply
rejection of Ho.

 It is thus similar to the t test - in fact ANOVA on 2

groups is equivalent to a t test [F = t2 ]
One way ANOVA’s limitations

 This technique is only applicable when there is one

treatment used.

 Note that this single treatment can have 3, 4,… ,many

levels.
 Thus nutrition trial on children weight gain with 4
different feeding styles could be analyzed this way,
but a trial of BOTH nutrition and mothers health
status could not
119

Thank You!