Biostatistics I: Basic for Public Health

Lecture No.: KUI 6111

Starting Date: 01/09/2017

Statistical Inference:
Hypothesis Testing for Continues Data
Module: 5

Copyright © 2017, S.A. Wilopo, Department of Biostataistics,

Epidemiology, and Population Health
Faculty of Medicine, Gadjah Mada University, Yogyakarta, Indonesia
Contents
1 Learning objectives 1

2 Activities 1

3 Exercise in Class 2
3.1 Hypothesis Tests: I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3.2 Hypothesis Tests II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.3 Hypothesis Tests: III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

4 Homework: 7
4.1 Critical Appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Comparing Two Groups for Continues Data using STATA . . . . . . . . . . . 8

5 References: 9
5.1 Articles for Critical Appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Required Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.3 Suggested Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

6 Outputs 10

7 Log sheet: 10

i
1 Learning objectives
Upon completion of the course unit, students should be able to:

1. Make decisions concerning unknown population parameters, based on observed sam-

ple data with continues measure

2. apply concepts of statistical inference on interval estimation for continues data

3. describe relationships between hypothesis testing and interval estimation

4. apply concepts of statistical inference on hypothesis testing using parametric and

non-parametric methods for continues data

5. appraise published research which used hypothesis testing for continues data

2 Activities
1. Discussion: Statistical Inference – Hypothesis Testing

2. Laboratory session:

(a) Hypothesis testing and types of errors (type I and II)

(b) P values and power
(c) Interval estimation for continues data
(d) the hypothesis testing for the difference in means between two groups
(e) Hypothesis testing for continues data – Independent and paired test
(f) Non-parametric tests for continues data

3. Homeworks:

(a) Critical appraisal on comparing means of two groups

(b) Calculating the difference in means between two groups using parametric and
non-parametric
(c) Reading and presenting computer outputs on the difference in means between
two groups using parametric and non-parametric

1
3 Exercise in Class
3.1 Hypothesis Tests: I
1. The level of significance, alpha, is the probability of

a) rejecting a false null hypothesis

b) accepting a true null hypothesis
c) rejecting a true null hypothesis
d) accepting a false null hypothesis

2. If an investigator rejects the null hypothesis

a) s/he has committed a Type II error

b) s/he has committed a Type I error
c) s/he has committed no error
d) s/he may have committed a Type II error
1) a, b, and c
2) a and c
3) b and d
4) d only
5) none of the above

3. A Type II error is

a) the probability that the null hypothesis is true

b) the probability that the null hypothesis is false
c) made if the null hypothesis is accepted when it is false
d) made if the null hypothesis is rejected when it is true
e) none of the above

4. The p-value is

a) the probability of the null hypothesis being true

b) the probability of the null hypothesis being false

2
 c) the probability of the test statistic or any more extreme results, assuming the
null hypothesis is true
d) the probability of the test statistic or any more extreme results, assuming the
null hypothesis is false
e) none of the above

5. A 95% confidence interval for the mean cholesterol level of adults over 65 years of
age is (198, 208) mg/dl. The mean cholesterol level for adults 40-60 years of age is
190 mg/dl. If a two-sided hypothesis test of Ho: µ=190 mg/dl were performed, we
would:

a) accept Ho at the 5% significance level

b) reject Ho at the 5% significance level
c) accept Ho at the 1% significance level
d) reject Ho at the 1% significance level
e) can’t tell

6. The 5% level of significance means:

a) we’re taking a 5% risk of misstating the null hypothesis

b) we’re taking a 5% risk that our sample is unrepresentative if the null hypothesis
is true
c) we’re taking a 5% risk of getting an unrepresentative sample if the alternative
hypothesis is true
d) we’re taking a 5% risk of making a wrong decision, regardless of which hypoth-
esis is true

7. Iron-deficiency anemia is an important nutritional health problem in Indonesia. A

dietary assessment was performed in 51 children of 9 to 11 year-old males whose
family were below the poverty line. The mean daily iron intake among these children
was found to be 12.50 mg with a standard deviation of 4.75 mg. Suppose that
the mean daily iron intake among a large population of 9-11 year-old boys from
all income strata is 14.44 mg. We wish to test if the mean iron intake among the
low-income boys is different from that of the boys in the general population. State
the hypothesis that can be used to consider this question.

3
 a) H0 : µ = 12.50 versus H1 : µ 6=12.50
b) H0 : µ = 14.44 versus H1 : µ = 12.50
c) H0 : µ = 14.44 versus H1 : µ < 14.40
d) H0 : µ = 14.44 versus H1 : µ 6=14.44
e) H0 : µ < 14.44 versus H1 : µ = 14.44

3.2 Hypothesis Tests II

1. A study was done to determine the effectiveness of an instruction booklet in improv-
ing nurses’ knowledge of testing for glycosuria in diabetes mellitus. A sample of 12
nurses was given a pretest prior to reading the booklet. After reading the booklet,
the same nurses were given a post-test.

Questions:

a) State the null hypothesis for determining if the instruction booklet was effective.
b) State the appropriate alternative hypothesis.
c) The appropriate t-statistic is:
1) Paired t-test
2) 2-sample t-test
d) The degrees of freedom associated with this test statistic are .

2. One method for assessing the effectiveness of a drug is to note its concentration in
blood and/or urine samples at certain periods of time after giving the drug. Suppose
we wish to compare the concentrations of two types of aspirin in urine specimens at
one time and measure the 1-hour urine concentration. One week later, after the first
aspirin has presumable been cleared from the system, we give the same dosage of the
other aspirin to the same person and note the 1-hour urine concentration. Since the
order of giving the drugs may affect the results, we use a table of random numbers to
decide which of the two types of aspirin to give first. We perform the experiment on
10 people. What statistical procedure would be appropriate for comparing the two
types of aspirin?

3. A study was performed in 1086 to relate the use of oral contraceptives with the levels
of various lipid fractions in a groups 163 non-pregnant, pre-menopausal women ages

4
 21-39. The serum cholesterol among 66 current users of oral contraceptives was
201 + 37 (mg/dl) (Mean + Standard deviation), whereas for 97 nonusers it was
193 + 37 mg/dl. What statistical procedure would be appropriate for determining if
cholesterol levels are affected by the oral contraceptive use?

4. In a pediatric clinic a study is carried out to see how effective aspirin is in reducing
temperature. Twelve 5-year-old girls suffering from influenza had their temperatures
taken immediately before and 1 hour after administration of aspirin. What is the ap-
propriate statistical procedure for determining if aspirin is reducing the temperature?
State the appropriate null and alternative hypothesis.

5. An investigator wishes to determine if sitting upright in a chair versus lying down on

a bed will affect a person’s blood pressure. The investigator decides to use each of
10 patients as his or her own control and collects systolic blood pressure data in both
the sitting and lying positions. What statistical procedure would be appropriate to
determine the effect of position on blood pressure?

3.3 Hypothesis Tests: III

1. A clinical trial was designed to test a drug that was believed to decrease blood-
clotting time. Forty subjects were selected and randomized to yield two groups,
each with n=20. One group was given the drug and the other group was given a
placebo, and served as a control. The mean clotting time, given in minutes, for the
drug treatment group is 4.90 minutes with variance of 10.24 minutes squared. The
mean clotting time for the control group is 7.45 and the variance is 12.96 minutes
squared.

Questions:

a) State the null hypothesis to test differences between the treatment and control
groups.
b) State the appropriate alternative hypothesis.
c) Using the above results, set up the appropriate test.

2. A class experiment in pharmacology consisted of distributing packets of instant coffee

to students. The contents of the packet were to be mixed with hot water and drank

5
 shortly before bedtime. The student received packets on two occasions: one time
the packet contained a placebo and the other time it contained coffee with caffeine.
Among other measurements, the students took their pulse rates (in beats per minute)
before consuming the instant coffee or placebo and then again afterward.
The students were classified as to whether they were coffee drinkers (those who
usually consumed two cups or more per day) or non-coffee drinkers (those who
usually consumed one or fewer cups per day). The results for 65 non-coffee drinkers
and 85 coffee drinkers are given below:

65 Non-Coffee Drinkers 85 Coffe Drinkers

Caffein Placebo Caffein Placebo
N 50 15 44 41
Mean 4.1 0.9 4.9 2

Questions:

a) State the null and alternative hypothesis for determining if the response is the
same among Coffee Drinkers.
b) What is the appropriate statistical procedure to test the null hypothesis in (a)?
c) What are the degrees of freedom associated with the test statistic in (b)?
d) State the null and alternative hypotheses to determine if the caffeine response
is the same between Non-Coffee and Coffee Drinkers.
e) What is the appropriate statistic procedure to test the null hypothesis in (d)?
f) What are the degrees of freedom associated with the test statistic in (e)?

3. A recent study attempted to compare the working environment in offices where smok-
ing was permitted with that in offices where smoking was not permitted. Measure-
ments were made of carbon monoxide (CO) at 1:20 pm in 40 work areas. Where
smoking was permitted, the mean CO=11.6 parts per million (ppm) and the stan-
dard deviation CO=7.3 ppm. Where smoking was banned, the mean CO=6.9 ppm
and the standard deviation CO=2.7 ppm. What statistical procedure would be ap-
propriate to see whether or not the mean CO is different in the two types of working
environments?

6
4 Homework:
4.1 Critical Appraisal
Read article with the following title:

Rerksuppaphol, S., and Rerksuppaphol, L. (2016). Effect of zinc plus multivitamin

supplementation on growth in school children. Pediatrics International, 58(11), 1193-
1199. doi:10.1111/ped.13011

or

Vizheh, M., Pakgohar, M., Babaei, G., & Ramezanzadeh, F. (2013). Effect of coun-
seling on quality of marital relationship of infertile couples: a randomized, con-
trolled trial (RCT) study. Archives of Gynecology and Obstetrics, 287:3, 583-589.
doi:10.1007/s00404-012-2595-9

or

Flores, G., Lin, H., Walker, C., Lee, M., Currie, J. M., Allgeyer, R., . . . Massey,
K. (2017). The health and healthcare impact of providing insurance coverage to
uninsured children: A prospective observational study. BMC Public Health, 1: 553.
doi:10.1186/s12889-017-4363-z
In those articles, authors use descriptive analyses and hypothesis testing. Please
answer the following questions:

(a) What types of statistical techniques have been shown in this article?
(b) Can you present some of the descriptive analysis into graphs? For example
comparing baseline data on stunting among treatment and control groups from
table 1 of article by Rerksuppaphol al. or table 1 from article by Vizheh et al. or
table 3 on variable "parental out-of-pocket costs of pediatric care" from article
by et Flores al. Please interpret!
(c) Please write-up the results of descriptive analyses used in this publication pre-
sented in the table 2.
(d) Identify what are hypothesis testing given to reader in this publication? Are
there pair t-test and independent t-test reported in this publication? Why?

7
 (e) Please write up the formal hypothesis statements in this article using two or
three results of the study available in tables (data) or in the text. For example:
state Ho and Ha as well as the critical value used in this publication (decision to
accept or reject the Ho ).

4.2 Comparing Two Groups for Continues Data using STATA

1. Let consider fem.dat in the previous exercise. Please conduct test on weight gain
among who have thought recently about ending their life or not. Having found no
strong evidence that the assumptions of the t-test are not valid for weight gain, we
can proceed to apply a t-test. Note that our previous test for equal variances is only
appropriate if the variable may be assumed to be normally distributed in each popu-
lation.

a. What is the difference in means between two groups. Does it clinically or prac-
tically important? Why?
b. What are 95% confidence intervals of the mean of each group?
c. What is a 95% confidence interval of the difference in means between two
groups?
d. What is the two-tailed p value for the difference in means between two groups?
e. Please summarize your analysis for the difference in means between two groups!
Present your result as part of your paper which is going to be submitted in the
journal.

2. Suppose we are conducting a study of the effectiveness of lithium therapy for manic-
depressive patients. The study is carried out at two different centers, and we want
to determine if the patient populations are comparable at baseline. A selfrating ques-
tionnaire about their general psychological wellbeing is administered to the prospec-
tive patients at the two different centers. The outcome measure on the questionnaire
is a four-category scale: (1) = feel good; (2) = usually feel good, once in a while
feel nervous; (3) = feel nervous half the time; (4) = usually feel nervous. Suppose
the data at the two different centers are as follows:
Center I 3, 4, 1, 1, 3, 2, 3, 4, 4, 3, 2, 4, 4, 4
Center II 1, 2, 1, 3, 2, 4, 1, 2, 1, 3, 1, 2, 2, 2, 1, 3

8
 Questions:

(a) What type of data does this type of scale represent?

(b) Why might a parametric test not be useful with this type of data? Can you test
assumption of the requirement for conducting t-test.
(c) Assess if there is any significant difference in the responses of the two patient
populations using a non-parametric test.
(d) Write up your analysis in the form of paragraph as part of your paper which is
ready to be submitted to the journal.

5 References:
5.1 Articles for Critical Appraisal
1. Rerksuppaphol, S., and Rerksuppaphol, L. (2016). Effect of zinc plus multivitamin
supplementation on growth in school children. Pediatrics International, 58(11), 1193-
1199. doi:10.1111/ped.13011

2. Vizheh, M., Pakgohar, M., Babaei, G., & Ramezanzadeh, F. (2013). Effect of coun-
seling on quality of marital relationship of infertile couples: a randomized, con-
trolled trial (RCT) study. Archives of Gynecology and Obstetrics, 287:3, 583-589.
doi:10.1007/s00404-012-2595-9

3. Flores, G., Lin, H., Walker, C., Lee, M., Currie, J. M., Allgeyer, R., . . . Massey,
K. (2017). The health and healthcare impact of providing insurance coverage to
uninsured children: A prospective observational study. BMC Public Health, 1: 553.
doi:10.1186/s12889-017-4363-z

5.2 Required Reading

1. Whitley E and Ball J. Statistics review 3: Hypothesis testing and P values. J. Critical
Care 2002; 6(3):222-225. Available online http://ccforum.com/content/6/
3/222

2. Whitley E and Ball J. Statistics review 5: Comparing means. J. Critical Care 2002;
6(6):425-428. Available online http://ccforum.com/content/6/5/424

9
 3. Whitley E and Ball J. Statistics review 5: Non-parametric method. J. Critical Care
2002; 6(6):509-513.Available online http://ccforum.com/content/6/6/509

5.3 Suggested Reading

1. Rosner, B. Hypothesis Testing: One sample inference. Chapter 8. Exercise of Funda-
mentals of Biostatistics, 5th ed. Belmont, CA: Duxbury Press, 2016; pp: 204-268

2. Rosner, B. Hypothesis Testing: Two sample inference. Chapter 8. Exercise of Funda-

mentals of Biostatistics, 7th ed. Belmont, CA: Duxbury Press, 2016; pp: 269-326

3. Rosner, B. Non-Parametric Method. Chapter 9 Exercise of Fundamentals of Biostatistics,

8th ed. Belmont, CA: Duxbury Press, 2015; pp: 237-351

6 Outputs
1. Discussion on hypothesis testing for continues data

2. Performing hypothesis testing for continues data: H0 and H A

3. Reading and understanding articles using t-test (paired and unpaired groups)

4. Understanding the use of non-parametric test

7 Log sheet:

No Activities Date Signature

1 Discussion on the concepts of hypothesis testing for
parametric and non-parametric test
2 Laboratory session: Hypothesis testing z and t-test for
continues data
3 Homework: Hypothesis Testing non-parametric test
for non-normal of continues distribution

Score : ____________________

10
 Instructor,

11