Sei sulla pagina 1di 29

Learning Module No.

3
T-Test of two independent samples

Situation

The Dean of the College of Arts and Sciences of the university was amazed to
find out that there were more female students majoring in mathematics. Their
choice of majors was based on an aptitude test. Because of this observation, he
wanted to find out if gender has an influence on the achievement of the
mathematics among the math major students. He got hold of the test results in basic
math and compared the mean scores between male and female students. Test the
hypothesis of no significant difference between the male and female score
(achievement) in mathematics at =0.05 using the data below.

Male Female
n 45 75
Mean 67.4 71.5
SD 12.3 13.8

1. Problem statement

Is there a significant difference in the basic math achievement between male


and female students majoring in mathematics?

2. Hypotheses (You need to state only one Ho and one Ha.)

Ho: There is no significant difference in the basic math achievement


between male and female students majoring in mathematics. (Ho: μ1 = μ2
[μ1 – male students and μ2 – female students])

Ha: There is a significant difference in the basic math achievement between


male and female students majoring in mathematics. (Ha: μ1 ≠ μ2, non-
directional two tailed hypothesis)
OR (Take note! Only one Ha is required.)

Ha: The female students have significantly higher achievement in basic math
than male students. (Ha: μ2 > μ1, directional or one tailed hypothesis)

3. Choice of test statistic and 


T-test of two independent samples,  = 0.05

X1 – X2
T-Value =
(SD1)² + (SD2)²
n1 n2

Where:
X1 = mean of first group X2 = mean of the second group
D1 = SD of first group SD2 = SD of second group
n1 = number of cases of first group n2 = number of cases of second group

4. Computation (Run the data either in the Minitab or in the SPSS.)

T-value = -1.69
P-value = 0.095

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value


Finding:  (0.05) < p-value (0.095)

6. Decision: Do not reject the null hypothesis.


7. Interpretation and analysis

There is no significant difference in the mean scores between male and female
students in mathematics. The performance of both gender groups in the
mathematics test is not comparable (meaning the mean scores of the two
groups are statistically equal).

8. Implications

The non-comparable achievement may be attributed to the same learning


environment of the students. They were with the same mathematics teachers.
They were admitted to the school by passing the same entrance examination;
hence, it can be assumed that they are of similar learning potentials. Even if
there were more female students who chose mathematics as their major, yet it
could not be said that they are better than their male counterparts in this subject
area.

9. Conclusion

Gender has no influence on the mathematics achievement of the students. It


can be said that males achieve as much as the female students in mathematics.

Classroom Exercise

Two photocopying machines were observed in a span of five years to find out
which brand was better. An investigation on the number of repair times was
recorded and the data are shown below.

Equipment Number of Repair Times (in minutes)


Brand Repair Jobs Mean SD

A 60 84.2 19.4
B 60 91.6 18.8

Test at the 0.05 level of significance whether the difference between these
two sample means repair time is significant.
Learning Module No. 4
T–test of two correlated samples or
T-test for paired/matched samples

This test statistic is commonly used in experimental research, when an intervention


or treatment is applied such as new method of teaching to improve students’ performance
using information technology (IT), a counseling program aimed at reducing stress level of
patients or a new weight lose exercise program. Mann (2004) stressed that two samples
are said to be paired or matched samples when for each data values collected from one
sample, there is a corresponding data value collected from the second sample, and both
these data values are collected from the same source. Any significant change
(improvement or reduction) in the measured variable after the treatment is attributed to
the effectiveness of the applied intervention. The data measurement required for this test
is either interval or ratio.

Situation

A certain Music Teacher of City Central School wanted to find out if her new method
of teaching music was effective. She administered a test in Music in June and administered
the same test at the end of the first grading period. The new method called Kudali
approach is teaching music using indigenous musical instruments made from bamboos or
wood. Music concepts were introduced using indigenous concepts of music from our
ancestors or from several ethnic groups. Fifteen pupils were selected at random. Test the
hypothesis of no significant improvement in the pupils' scores from the pre-test to the
post-test using the following results:

Pupil Test in June Test in August


(Pretest) (Post test)
1 7 20
2 10 11
3 11 14
4 8 16
5 12 18
6 6 16
7 10 10
continued...
Pupil Test in June Test in August
(Pretest) (Post test)
8 12 10
9 9 20
10 8 19
11 11 20
12 13 19
13 8 18
14 10 17
15 11 16

1. Problem statement

Is there a significant improvement of the students’ music scores from the pre-test to
the post-test?

OR (Take note! Only one problem statement is required.)

Is there a significant pre-post mean gain of the student scores in music?

2. Hypotheses

Ho: There is no significant improvement in the students’ scores from the pre-test to
the post-test.

OR (Take note! Only one Ho is required.)

There is no significant pre-post mean gain of the student scores in music.

Ha: There is a significant improvement in the students’ scores from the pre-test to
the post-test.

OR (Take note! Only one Ha is required.)

There is a significant pre-post mean gain of the student scores in music.


3. Choice of test statistic and : T-test of correlated samples,  = 0.05

Xd where:
T-Value = Xd = mean of the difference
SDd SDd = SD of the difference
N N = number of cases

Note: mean difference = post – pre or after – before

4. Computation (Run the data either in the Minitab or in the SPSS.)

T-value = 5.70
P-value = 0.00

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value


Finding:  (0.05) > p-value (0.00)

6. Decision: Reject the null hypothesis.

7. Interpretation and analysis

There was a significant improvement of the students’ scores in music from the pre-
test to the posttest. Generally, the students’ scores increased in the test administered
in August.

8. Implications

Students enjoyed the use of indigenous musical instruments. The concepts were new
so they were excited to be exposed to the music of our ancestors and the different
ethnic groups.
9. Conclusion

The Kudali approach was an effective approach in the teaching of music among the
Grade Six pupils.

Classroom Exercise

Find out the effect of daily exercise in the reducing program of 20 ladies using their
weights before and after the program at  = 0.05.

Case Before After Case Before After


1 142 132 11 126 120
2 130 126 12 135 133
3 128 125 13 142 132
4 118 114 14 130 126
5 126 120 15 128 125
6 135 133 16 130 133
7 142 132 17 128 132
8 130 126 18 118 126
9 128 125 19 126 125
10 118 114 20 135 114

The Chi-Square Distribution


(Mann, 2004; Doane and Seward, 2009)

The chi-square distribution has only one parameter (v)


called the degrees of freedom (df). Its shape depends on the
number of df, the curve is skewed for very small df, and
drastically changes as the df increases. The chi-square
distribution curve is skewed to the right of the vertical axis for
small df and becomes symmetric for large df which assumes
nonnegative values only. As the df increases, the chi-square
distribution approaches a normal distribution. The mean of a
chi-square distribution is its df. The mode is df - 2 and the
median is approximately df - 0.7.
Learning Module No. 5
Chi-square test of goodness of fit

The chi-square goodness-of-fit test is always a right-tailed test. Whether or not the
Ho is rejected depends on how much the observed (Fo) and expected (Fe) frequencies
significantly differ from each other. It should be noted that to make a goodness-of-fit test,
the sample size should be large enough so that the expected frequency for each category
is at least five. When there is a category with Fe less than five, either increase the sample
size or combine two or more categories to make each Fe at least five. For more than the
2x2 contingency table, the chi-square test should not be used if more than 20 percent of
the expected frequencies are less than five. The chi-square test statistic can only be
applied to either nominal or ordinal (categorical) data.

There are two situations of expected frequencies (Fe) where you are going to use
and apply the chi-square test statistic, the first is with equal expected frequency and the
second one is with unequal expected frequency. For the latter, you compare the observed
and the predicted frequencies where equal frequency cannot be expected. Here, you test
the null hypothesis that the observed frequencies for an experiment/research follow a
certain pattern or theoretical distribution.

5.1 Situation 1: Equal Expected Frequency

A random sample of 120 high school teachers was drawn from a Cebuano
population. The sample teachers were asked to rate the job satisfaction items using the
Likert scale. The self-perceptions made by the subjects were categorized into three
satisfaction levels. The resulting data of teachers’ degree of job satisfaction is presented
below.

Degree of Job Satisfaction n Expected Frequency Difference


(absolute)

Very Satisfied 42 40 2
Satisfied 75 40 35
Less Satisfied 3 40 37
Total 120
Test the null hypothesis that the relative frequency of teachers who are very
satisfied is the same as those who are satisfied or less satisfied. What do the results
indicate?

1. Problem statement

Is there a significant difference between the observed frequency and expected


frequency of the teachers’ job satisfaction groups?

OR (Take note! Only one problem statement is required.)

Does the sampling distribution differ significantly from the hypothesized population
distribution (equal expected frequency)?

2. Hypotheses (you need to state only one Ho and one Ha)

Ho: There is no significant difference between the observed and expected frequency
of the teachers’ job satisfaction groups?
(Ho: fo = fe or Ho: fo – fe = 0)

OR (Take note! Only one Ho is required.)

Ho: The sampling distribution did not differ significantly from the hypothesized
population distribution. (Ho: fo = fe or Ho: fo – fe = 0)

Ha: There is a significant difference between the observed frequency and expected
frequency of the teachers’ job satisfaction groups. (Ha: fo ≠ fe)

OR (Take note! Only one Ha is required.)

Ha: The sampling distribution significantly differs from the hypothesized population
distribution. (Ho: fo ≠ fe or Ho: fo – fe ≠ 0)
3. Choice of Test Statistic:
Chi-square test of equal expected frequency,  = 0.05.

(Fo – Fe)² Where: Fo = observed frequency


2
X -value = ∑ Fe = expected frequency
Fe ∑ = summation

4. Computation (Run the data either in the Minitab or in the SPSS)

X2-value = 64.95
P-value = 0.00

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value


Finding:  (0.05) > p-value (0.00)

6. Decision: Reject the null hypothesis.

7. Interpretation and analysis

There is a significant difference between the observed and the expected frequency.
The sampling distribution differs significantly from the hypothesized equal
distribution. There were more satisfied teachers and very satisfied teachers than
expected.
8. Implications Ranges for the Weighted Mean
For three levels
2.34 – 3.00 – Very Satisfied (High)
Teachers being satisfied on their job could be 1.67 – 2.33 – Satisfied (Average)
attributed to the school environment and 1.00 – 1.66 – Less Satisfied (Low)
the collegial peer relationship. The
administration may be supportive of their For five levels
professional needs and provide assistance in 4.21 – 5.00 – Very High
3.41 – 4.20 – High
solving their problems.
2.61 – 3.40 – Average
1.81 – 2.60 – Low

9. Conclusion

On the average, the teachers’ ratings were on the satisfied level. In other words, based
on the findings considering all of teachers’ rating, they were satisfied with their job.

Based on the computation of the weighed mean.

∑fw
WM =
n

(42x3) + (75x2) + (3x1)


WM =
120

WM = 2.33

Classroom Exercise

Eighty-two High School Principals in Cebu Province were asked about their opinions
on the desirability of integrating Sex Education in the HS curriculum. Thirty-one of the
sample School Administrators agree, 22 tend to agree, 18 tend to disagree while 11
disagree. Is there a significant difference in the opinion of HS Principals regarding the
integration of Sex Education in the HS curriculum? Test at 0.05 level of significance.
5.1 Situation 2: Unequal Expected Frequency

On divorce issue in the Philippines, it can never be expected that there will be equal
number of those who agree and disagree. Since the Philippines is predominantly Christian,
it is expected that there will be more women who will disagree than those who will agree.
Thus an expected 30–70 ratio is hypothesized in favor for the no response.

In a sample of 150 mothers, test the hypothesis of no significant difference between


the observed and hypothesized distribution of population using the data below:

Opinion n
Yes 42
No 108

1. Problem

Is there a significant difference between the observed and the expected frequency
(30–70 distribution) among the mothers’ opinions on legalizing divorce in the
Philippines?

2. Hypotheses (You need to state only one Ho and one Ha.)

Ho: There is no significant difference between the observed and the expected
frequency (30–70 distribution) among the mothers’ opinions legalizing divorce in
the Philippines.

OR (Take note! Only one Ha is required.)

Ho: There is no significant difference between women who disagree and agree.

Ha: There is a significant difference between the observed and the expected
frequency (30–70 distribution) among the mothers’ opinions legalizing divorce in
the Philippines.
OR (Take note! Only one Ha is required.)

Ha: There are significantly more women who disagree than those who agree.

3. Choice of Test Statistic:


Chi-square test of unequal expected frequency,  = 0.05

(Fo – Fe)² Where: Fo = observed frequency


2
X -value= ∑ Fe = expected frequency
Fe ∑ = summation

4. Computation (Run the data either in the Minitab or in the SPSS.)

X2-value = 0.286
P-value = 0.593

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value


Finding:  (0.05) < p-value (0.593)

6. Decision: Do not reject the null hypothesis.

7. Interpretation and analysis

There was no significant difference between the mothers’ opinion on legalizing divorce
in the Philippines and the theorized 30 – 70 distribution. This means that the sample
population come up with the expected 30 – 70 distribution in favor of the “no” group.
As expected, there were few mothers who agreed to legalize divorce in the country.
8. Implications

Most of the mothers’ negative opinions may be attributed to religious beliefs that only
God can separate what He has joined in marriage.

9. Conclusion

Filipino mothers do not like to legalize divorce in the Philippines.

Classroom Exercise

A mixing nut machine was designed to mix peanuts, hazelnuts, cashews and pecans
following a 50:20:20:10 grams ratio. Twenty-five randomly selected 500 grams mixed-nut
cans revealed, on the average, 269 grams of peanuts, 112 grams of hazelnuts, 74 grams of
cashews, and 45 grams of pecans. Test at 5 percent level of significance whether the
machine is fitted to produce the designed ratio based on the given sample data.
Learning Module No. 6
Chi-square test of independence

Relationship between two variables with nominal data is determined using the chi-
square test of independence. Significant chi-square value means a presence of
relationship. The degree of relationship could be computed using the coefficient of
contingency. Here one tests the Ho that the two attributes (characteristics) of the elements
of a given population are not related (independent) against the Ha that two characteristics
are related. (Mann, 2004; Doane and Seward, 2009)

6.1 Situation 1: 2x2 contingency table

Forty-four high school students were categorized as drinkers and non-drinkers (See
data below). It was hypothesized that the proportions of drinkers among college bound
and non-college bound high school students are the same as the proportions of non-
drinkers who are college bound and not college bound. Using the data, find out if status of
drinking and college disposition are related at 0.01 level of significance.

Category Go to college Not go to college


Drinkers 17 (A) 7 (B)
Non-drinkers 8 (C) 12 (D)

1. Problem statement

Is there a significant relationship between status of drinking and college disposition?

2. Hypotheses

Ho: There is no significant relationship between status of drinking and college


disposition.
Ha: There is a significant relationship between status of drinking and college
disposition.

3. Choice of Test Statistic: Chi-square test of independence samples,  =0.01

Formula for a 2 x 2 contingency table

N(AD – BC) ²
2
X -value =
(A+B) (C+D) (A+C) (B+D)

The assignment of letters in the 2x2 contingency table is as follows:

Status of College Disposition


drinking Going to college Not going to college
Drinkers 17 (A) 7 (B)
Non-drinkers 8 (C) 12 (D)

3. Computation (Run the data either in the Minitab or in the SPSS.)

X2-value = 4.227
P-value = 0.040

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value.


Finding:  (0.01) < p-value (0.040)

6. Decision: Do not reject the null hypothesis.

7. Interpretation and analysis

There is no significant relationship between the students’ status of drinking and their
college disposition. College disposition does not relate on their status of drinking. The
results indicate that the proportion of high school alcohol drinkers who will study in
college is not significantly different from those who will not study in college. This also
indicates that the proportions in the present samples are not significantly different
from the proportions in the population from where these subjects drawn from.

8. Implications

The desire of going/not going to college is not related to a drinker/non-drinker


behavior. Such behavior is not an indicator for an academic pursuit. Alcoholic drinkers,
who could be children of well-off or poor families, have equal chances to study college
as those who are non-alcoholic coming from rich and financially handicapped families.
Going or not going to college is a self-driven desire/ motivation from which being a
drinker/non-drinker has no bearing.

9. Conclusion

College disposition has no relationship with the status of drinking. There is


theoretically the same proportion of drinkers and non-drinkers alike who are college
and non-college bound.

Classroom Exercise

Two drugs were administered to two groups of randomly assigned patients to cure
the same disease. One hundred twenty patients were involved in the research with the
data given below revealing the number of patients who were cured and not cured by each
of the two drugs. Test at 0.01 level of significance whether the two drugs are similar (or
otherwise) in curing and not curing the patients.

Drug Potency of Drug


type Cured Not Cured
Drug 1 53 19
Drug 2 19 29

6.2 Situation 2: Beyond the 2x2 contingency table


A survey among victims of typhoon Ondoy in Metropolitan Manila was conducted
to evaluate their perception about the government's level of responsiveness on post-
calamity stricken areas. The table below reflects the responses of the affected citizenry. Is
there a significant relationship between gender and perceived level of government's
responsiveness among the typhoon victims? Test at 0.05 level of significance.

Highly Responsive Less


Gender Responsive Enough Responsive
Male 495 567 862
Female 536 698 842

1. Problem statement

Is there a significant relationship between gender and perceived level of government's


responsiveness among the typhoon victims?

2. Hypotheses

Ho: There is no significant relationship between gender and perceived level of


government's responsiveness among the typhoon victims.

Ha: There is a significant relationship between gender and perceived level of


government's responsiveness among the typhoon victims.

3. Choice of test statistic: Chi-square test of independence,  =0.05

(Fo – Fe)² Where: Fo = observed frequency


2
X -value = ∑ Fe = expected frequency
Fe ∑ = summation

4. Computations (Run the data either in the Minitab or in the SPSS.)


X2-value = 9.669
P-value = 0.008

5. Decision rule and finding

Decision Rule: Reject Ho if level of significance () > p-value.


Finding:  (0.05) > p-value (0.008)

6. Decision: Reject the null hypothesis.

7. Interpretation and analysis

There is a significant relationship between gender and perceived level of government


responsiveness on post-calamity stricken areas among the victims of typhoon Ondoy.
Gender has strong association with perceived level of responsiveness by the
government towards the affected citizenry. It matters being a male or a female in
terms of how one views government’s responsiveness to assist calamity victims. Based
on the data, a considerable proportion of males tend to perceive government as less
responsive.

8. Implications

Government’s responsiveness to assist and help calamity victims is a social


responsibility which can be perceived and viewed differently between male and female
citizenry. Government’s disaster coordinating agencies may also give priorities to
women as more vulnerable groups than their men counterparts in terms of calamity
assistance provisions.
9. Conclusion

Gender has bearing on the perception regarding government’s responsiveness to help


out typhoon victims. There is a higher tendency that males would perceive
government to be less responsive whereas females have higher tendency to judge
government’s responsiveness as just enough.

Classroom exercises

Mr. Juan Dela Cruz is a statistics Professor who teaches graduate statistics courses
to English and Social Sciences masters students. He has observed that many of his students
do not like the rigors of mathematical computing during class sessions especially when
they have to use calculators instead of computers. He fielded a questionnaire to examine
if students’ attitude has bearing on their midterm and final examination results in his
statistics subject. With the data he gathered as shown in the table below, test whether
attitude relates to statistics examination performance at 0.05 level of significance.

Test Performance
Attitude AA A BA
Positive 27 32 8
Negative 18 25 36
Legend: AA – above average
A – average
BA – below average

Learning Module No. 8


Pearson’s coefficient of correlation or
Pearson product moment correlation or
Pearson r

In statistics, the Pearson product-moment correlation coefficient (r) is a common


measure of the correlation between two variables X and Y. When measured in a population
the Pearson Product Moment correlation is designated by the Greek letter rho (ρ). When
computed in a sample, it is designated by the letter "r" and is sometimes called "Pearson's
r" (Doane and Seward, 2009).

Situation

A researcher wants to examine if work environment is related to the work efficiency


among CNU clerical employees. Twenty clerks were chosen at random to become the
respondents of the study. Their assessment of the work environment and efficiency ratings
were obtained as indicated below.

Clerk Work Environment Work Efficiency


1 74 93
2 75 90
3 90 82
4 68 86
5 80 85
6 72 79
7 90 86
8 68 85
9 80 79
10 72 93
11 74 90
12 75 82
13 90 93
14 68 90
15 80 82
16 72 86
17 90 85
18 68 79
19 80 93
20 72 90

It should be noted that the situation is given only for the sake of illustrating how to
compute and analyze a statistical problem which calls for Pearson r statistic. The 20-sample
observations are small that in actual research, this should be increased to at least 30
observations.

1. Problem statement
Does relationship exist between the work environment and efficiency ratings of the
clerks in CNU?

2. Hypotheses

Ho: There is no significant relationship between work environment and efficiency


ratings of the clerks in CNU.

Ha: There is a significant relationship between work environment and efficiency


ratings of the clerks in CNU.

3. Choice of test statistic : Pearson product moment of correlation,  = 0.05

∑ xy
r-value =
(SDx )(SDy)(N)

where: ∑xy = sum of the products of x and y deviations


SDx = standard deviation of x variable
SDy = standard deviation of y variable
N = number of cases

4. Computation (Run the data either in the Minitab or in the SPSS.)

r-value = -0.007
P-value = 0.976

5. Decision Rule and Finding

Decision Rule: Reject Ho if level of significance () > p-value


Finding:  (0.05) < p-value (0.976)
6. Decision: Do not reject the null hypothesis.

7. Interpretation and analysis

There is no significant relationship between the work environment and the efficiency
ratings of the clerks in CNU. Clerks with better work environment may or may not be
efficient workers.

8. Implications

Efficiency of work is a result of one’s capability and commitment to job, so even with
a marginalized work environment still a person can work efficiently if he/she has the
capabilities and the desire to work better.

9. Conclusion

A good working environment could not guarantee efficiency of work.

Classroom Exercise

Shown on the table below are the average daily consumption of saturated fat (in
grams) and the cholesterol level (in milligrams per hundred millimeters) of 10 randomly
chosen sample of patients at the Cebu City Medical Center. Is there a significant correlation
between fat consumption and cholesterol level? Test at 0.05 level of significance.

Fat consumption 65 78 60 44 53 68 87 46 55 73
Cholesterol level 190 225 205 175 180 214 245 160 210 215

You should take note that even if these variables (fat consumption and cholesterol
level) may be highly correlated, it is not a sufficient proof of causation. To be able to
establish causation, all predicting variables for cholesterol level must be considered like
health, age, nutritional status and stress level, among others. To conclude causation, the
causal variables must precede the variable it causes, and several conditions must be met
(e.g., reversibility, strength, exposure response, etc.).
The F-Distribution
(Doane and Seward, 2009; Lind, Marchal and Wathen, 2006)

The F-Distribution is named in honor of Sir Ronald A. Fisher (1890-1962), one of the
founders of modern day statistics. This probability distribution is used to test whether two
(or more) samples are from the same populations having equal variances. It is also applied
when you want to compare several population means simultaneously. This simultaneous
comparison of several population means is called analysis of variance (ANOVA).

ANOVA, as a hypothesis-testing technique is used to test the equality of two or more


population (or treatment) means by examining the variances of samples that are taken.
The term treatment is used to identify the different populations being examined. ANOVA
is based on comparing the variance (or variation) between the data samples to variation
within each particular sample. If the between variation is much larger than the within
variation, the means of different samples will not be equal. If the between and within
variations are approximately the same size, then there will be no significant difference
between sample means. It then allows one to determine whether the differences between
the samples are simply due to random error (sampling errors) or whether there are
systematic treatment effects that cause the mean in one group to differ from the mean in
another.

To use ANOVA, the following assumptions have to be met:

1. All populations involved follow a normal distribution;


2. All populations have the same variance (or standard deviation); and
3. The samples are randomly selected and independent of one another.

When it is determined that differences exist among treatment means, post hoc
range tests and pairwise multiple comparisons are further employed to uncover which
among the treatment means differ from each other. Range tests ascertain homogeneous
subsets of means that are not different from each other. Using the Minitab to run the
ANOVA and opting to use the Tukey’s Honestly Significant Difference (HSD), the pairwise
multiple comparisons test the difference between each pair of means and yield a grouping
where letters denote significantly different group means at a given alpha level (i.e. α =
0.05).

Learning Module No. 10


F-Test: One Way Analysis of Variance (ANOVA)
(Test of mean difference for 3 or more samples)

To test whether the independent variable explains a significant proportion of the


variation in Y, one needs to compare the explained (due to treatment) and unexplained
(due to error) variation. Note that the F distribution describes the ratio of two variances,
so that the ANOVA test statistic is the F test statistic which is the ratio of the variance due
to treatments to the variances due to error. The one-way ANOVA table is outlined below.

Source of Degree of Sum of Mean


P-
Variations (SV) Freedom Squares Squares F-Value
Value
(df) (SS) (MS)
Treatment Treatment Treatment Treatment MS
t-1
(Bet. Columns) SS MS Error MS
Expt’l .Error Error SS Error MS
t (r -1)
(Within Columns)
Total (t)(r) - 1 Total SS

Where: r = row df: r-1 = 3-1= 2


t = treatment (column)
df = degrees of freedom

Situation

Four groups of Cebu City Medical Center (CCMC) patients (3 person in each group)
were treated with different medications for the same ailments. The data below show the
number of average days it took the patients to recover from the start of ailment. Is there
enough evidence to conclude that there was significant difference (α = 0.01) in the number
of recovery days (hence significant difference in the effect of medications)?

Patient Group Medicol Biogesic Panadol Treuphadol


1 6 3 4 6
2 7 4 5 7
3 5 3 5 8

1. Problem Statement
Is there a significant difference in the mean number of recovery days among patient
groups?

OR (Take note! You are only one required to state one problem.)

Is there a significant effect of medication among treatments?

2. Hypotheses

Ho: There is no significant difference in the mean number of recovery days among
patient groups.

OR (Take note! Only one Ho is required.)

Ho: There is no significant difference between types of medications.

Ha: There is significant difference in the mean number of recovery days among patient
groups.

OR (Take note! Only one Ha is required.)

Ha: There is a significant difference between types of medications.

3. Choice of test statistic : F-Test (One way ANOVA),  = 0.01

4. Computation (Run the data either in the Minitab or in the SPSS.)

F-value = 11.46
P-value = 0.003
The resulting ANOVA table with computed values is shown below.

Degrees of Sum of Mean


Source of F-
Freedom (DF) Squares (SS) Squares P-Value
Variations (SV) Value
(MS)
Treatment
22.917 7.639 11.46** 0.003
(Between Columns) 3
Expt’l. Error
8 5.333 0.667
(Within Columns)
Total 11 28.250
** - highly significant at α = 0.01

5. Decision Rule and Finding

Decision Rule: Reject Ho if level of significance () > p-value.


Finding:  = 0.05 > p-value (0.003)

6. Decision: Reject the null hypothesis.

7. Interpretation and analysis

There is significant difference in the mean number of days of patient group recovery.
There is significant difference between types of medications as to their effects on
patients’ recovery.

Based on the post-hoc analysis employing the Tukey statistic, treatment means having
the same grouping letters do not significantly differ from each other.

Treatment n Mean Grouping


Treuphadol 3 7.0000 A
Medicol 3 6.0000 AB
Panadol 3 4.6667 BC
Biogesic 3 3.3333 C

Overall, the paracetamol treatment showed significant effect on the number of


recovery days among the patients. Treupadol (A) significantly differs from panadol (B)
and biogesic (C) in terms of its treatment effects. In like manner, treupadol and
medicol (both A) differs significantly from biogesic (C). Biogesic showed the fastest
mean number of days of recovery as compared to other paracetamol used in the
experiment.
Meanwhile, treupadol and medicol do not significantly differ from each other (having
the same group letter A), the same holds true between medicol and panadol (having
the same group letter B) and between panadol and biogesic (having the same group
letter C).

8. Implications

Although the treatments are of the same type of medicines which is paracetamol, their
effects in terms of number of recovery days are different for the same ailment of
patients. It could be that some patient groups are more receptive to paracetamol
treatment than the other patient groups.

9. Conclusion

There is enough evidence to conclude that there were differences in the number of
recovery days as affected by the treatment medications. The type of paracetamol has
effect on the number of recovery days of patients. Biogesic is the most effective
paracetamol with regard to its effect on the patients’ recovery from the ailments.

Classroom Exercise

A course on inferential statistics was taught in the CNU graduate school to three
different classes of students by three different professors. At the end of the semester, the
students were given the same departmental examination of 100 pts. The scores of
randomly selected students are given below. Is there any significant difference in the
scores between the three groups of students?

Class 1 Class 2 Class 3 Class 1 Class 2 Class 3


95 79 75 93 78 73
83 82 68 81 80 66
87 81 71 85 79 69
85 69 67 83 67 65
86 77 73 84 75 71
89 63 70 87 64 68
94 78 72 92 76 70
87 84 67 85 82 65
93 79 73 91 77 71
88 81 69 86 79 67

Potrebbero piacerti anche