Research Paper
1. Introduction.
In statistics, the Kolmogorov–Smirnov Test (K–S Test) is a
nonparametric test for the equality of continuous, one-dimensional
probability distributions that can be used to compare a sample with a
reference probability distribution (One-Sample K–S Test), or to compare
two samples (Two-Sample K–S test). The Kolmogorov–Smirnov statistic
quantifies a distance between the empirical distribution function of the
sample and the cumulative distribution function of the reference
distribution, or between the empirical distribution functions of two
samples. The null distribution of this statistic is calculated under the null
hypothesis that the samples are drawn from the same distribution (in the
two-sample case) or that the sample is drawn from the reference
distribution (in the one-sample case). In each case, the distributions
considered under the null hypothesis are continuous distributions but are
otherwise unrestricted. The Two-Sample K – S Test is one of the most
useful and general nonparametric methods for comparing two samples,
as it is sensitive to differences in both location and shape of the
empirical cumulative distribution functions of the two samples. The
Kolmogorov–Smirnov Test can be modified to serve as a goodness of fit
test. In the special case of testing for normality of the distribution,
samples are standardized and compared with a standard normal
distribution. This is equivalent to setting the mean and variance of the
reference distribution equal to the sample estimates, and it is known that
using these to define the specific reference distribution changes the null
distribution of the test statistic. Various studies have found that, even in
this corrected form, the test is less powerful for testing normality than the
Shapiro–Wilk test or Anderson–Darling test.
3. One-Sample T-Test
It is perhaps easiest to demonstrate the ideas and methods of the one-
sample t-test by working through an example. To reiterate, the one-
sample t-test compares the mean score of a sample to a known value,
usually the population mean (the average for the outcome of some
population of interest). The basic idea of the test is a comparison of the
average of the sample (observed average) and the population
(expected average), with an adjustment for the number of cases in the
sample and the standard deviation of the average. Working through an
Research Methodology For Management Decision
example can help to highlight the issues involved and demonstrate how
to conduct a t-test using actual data.
In the United States, mothers who live in poverty generally have babies
with lower birth weight than those who do not live in poverty. While the
average birth weight for babies born in the U.S. is approximately 3300
grams, the average birth weight for women living in poverty is 2800
grams.
Eliminating the linkage between poverty and low birth weight status has
been a prominent dimension of health policy for the past decade.
Recently, a local hospital introduced an innovative new prenatal care
program to reduce the number of low birth weight babies born in the
hospital. In the first year, 25 mothers, all of whom live in poverty,
participated in this program. Data drawn from hospital records reveals
that the babies born to these women had a birth weight of 3075 grams,
with a standard deviation of 500 grams.
Research Methodology For Management Decision
The question posed to you, the researcher, is whether this program has
been effective at improving the birth weights of babies born to poor
women.
a. Establish Hypotheses
In this case:
Extension Exercise
Are the birth weights of the babies born to the participants of the
prenatal care program significantly different from the average for the
overall national average?
• You hear that the average person sleeps 8 hours a day. You think
college students sleep less. You ask 10 college kids how long they
sleep on an average day.
• You get the data and the mean of sleep time is 6.5 hours.
• Is this luck? Did you happen to pick a group of light sleepers by
chance? Or do college kids really sleep less?
• What you are asking is whether college kids come from a
population separate from the rest of society. "Society" supposedly
being all the people who contributed to the mean of 8 hours a night
that we hear about.
The Big Idea: Does Your Group Come from a Different Population
than the One Specified?
Research Methodology For Management Decision
• The group comes from the same population and the mean varies
by chance. You just happened to pick a sample group with a
mean that misrepresents the population it came from. The group
isn't really different. When this happens you fail to reject the Null
Hypothesis - see the graphic on the next page.
The way to decide which is the case is with the one sample t-test.
You will compare the sample mean to the population mean and get an
estimate of the probability that the sample mean is different by chance.
How does the t-test work?
First, you will calculate the mean and standard deviation for your
group. A computer might actually do all of this for you.
Next:
• The t-test calculates if your sample mean is far in the tails of the
sampling distribution and far away from the population mean (m).
• Thus, the difference may be too big to be chance. If your group
was really from the null hypothesis distribution, then the difference
should be close to zero.
One-Sample t Test
The Data
Staff Publications
Member
1 7
2 3
3 44
4 4
5 21
6 11
7 10
8 3
9 0
10 6
11 0
12 1
13 2
14 15
15 15
16 0
17 28
Research Methodology For Management Decision
µ 10
11.52
in the alternative that UNE staff publish more than the national
average, our alternative hypothesis would be H1: m > 8 and
we'd be interested only in the probability of getting a sample
mean of 9.4. We would ignore evidence in the other direction
even if the deviation in the other direction is substantial. The
issue of whether to use two-tailed tests or one-tailed tests is a
thorny one. Experts around the world cannot agree. For our
purposes, we will always be using two-tailed tests. In the
School of Psychology at UNE, two-tailed tests are the standard
approach even when the research hypotheses are phrased in
a directional manner.
Call the total "area" under this curve and above the x-axis "1"
Research Methodology For Management Decision
to denote the fact that the probability is 1 that you will get a t
statistic of some value. The obtained t of 0.372 is marked on
this figure. This value of t corresponds to a sample mean of
9.4. The hashed proportion of the area under the t distribution
to the right of 0.372 represents the probability of finding a t
statistic equal to or greater than 0.372 if the null hypothesis is
true, or of finding a mean of 9.4 or greater if the null hypothesis
is true. The -0.372 is also of interest because we are doing a
two-tailed test. We want to know the probability of getting a t
statistic as large or larger than 0.372, regardless of the
direction of the difference between the sample mean and the
null hypothesised value of the population mean. Notice that
-0.372 corresponds to a sample mean of 6.6 (you can verify if
you want by plugging 6.6 into the t statistic formula above).
The hashed area under the t distribution to the left of -0.372
represents the probability of finding a t statistic equal to or less
than -0.372 if the null hypothesis is true, or of finding a sample
mean of 6.6 or less if the null hypothesis is true.
You can see that SPSS prints out all the information we could
have computed by hand, including the sample mean (Mean),
sample standard deviation, sample size. The standard error of
the sample mean (Std. of Mean), t statistic (under "t"), degrees
of freedom (df) and the all important p value are also printed.
The p value is listed under "Sig. (2-tailed)". In the present
example, the probability that a difference as large as the one
we have observed (or greater), given that the null hypothesis is
true, is equal to .729. Thus, it appears that what we have
observed is quite likely to occur if the null is true, and so we
have no grounds on which to reject Ho as a plausible
explanation of our data!
T-Test
Research Methodology For Management Decision
You can use the PROB20 program that was included with the
floppy disk that was sent to you to find p-values for t. It will
have installed to C:\202progs\prob20\Prob.exe, and you can
use File Manager (in Windows 3.1) or Explorer (in Win95) to
locate it. Double-clicking will bring up a DOS window with
PROB20 running. Use the space bar to shift between
distributions (Z, t, etc.) and tab to shift between windows. For
the preceding example, we see
find a sample mean that deviates at least 1.4 from the null
hypothesised value of 8 about 3 in 4 times (which is p = .75)
we take a random sample of size 5 from this population. The
obtained difference simply isn't sufficiently rare enough for us
to rule out "chance" or sampling error as the explanation for
the difference between the sample mean and the null
hypothesised value of the population mean of 8. So we
conclude that the psychology academics at UNE publish about
the same as the national average: 8 times per staff since 1974.
Had p been less than alpha (.05 here), we would have rejected
the null hypothesis. When the result leads us to reject the null
hypothesis, we claim the result is statistically significant.
hypothesis.
Example 2
The five steps of hypothesis testing for the one sample t-test
will be briefly presented for the problem of determining if the
average number of publications for the entire staff at UNE
School of Psychology differ from the national average of 8.
That is, all 17 observations now form a sample instead of a
population.
Ho: µ = 8.0
H1: µ 8.0
24.1617
22.6681
25.8187
25.5415
24.6176
24.0489
25.8860
26.1441
23.8761
24.4975
24.1747
24.7755
23.1442
Check for Stability
Our next move is to check the stability of the data. We recently read on
an excellent lean six sigma blog that this was really important. So, we
took our data and entered it into Minitab, which is a standard off the
shelf statistical software package, and constructed an Individual’s Chart
(I Chart) as shown below:-
Observation Individual Value
1 3 5 7 9 11 13 15 17 19
27
26
25
24
23
22
X=24.680
UCL=27.004
LCL=22.357
I Chart of Critical Dimension
While there seems to be a fair bit of common cause variation in our
process it looks as though the data is relatively stable and no trends
appear to be developing. Next up we need to check out this funky thing
called normality. Scroll down for more fun sports fans.
Check for Normality
Our next move was to determine whether or not the data were normally
distributed. To do this we used Minitab to help us study the “descriptive
statistics” in a graphical manner as shown below.
Research Methodology For Management Decision
23 24 25 26
Median
Mean
24.00 24.25 24.50 24.75 25.00 25.25 25.50
1st Quartile 24.021
Median 24.558
3rd Quartile 25.756
Maximum 26.144
24.212 25.148
24.051 25.607
0.761 1.461
A -Squared 0.57
P-Value 0.123
Mean 24.680
St Dev 1.000
Variance 1.000
Skewness -0.120643
Kurtosis -0.802125
N 20
Minimum 22.668
Anderson-Darling Normality Test
95% Confidence Interval for Mean
95% Confidence Interval for Median
95% Confidence Interval for St Dev
95% Confidence Intervals
Summary for Critical Dimension
Before looking at the numbers we looked at the shape of the distribution.
We noticed that it was not the prettiest bell curve we had ever seen but
also realized that with such as small sample size we may not want to
beat ourselves up too much. So, we took a look at the Anderson-Darling
Normality Test results in the top right hand corner. Our null hypothesis
(Ho) and alternate hypothesis (H1) for this normality test are as follows:
Ho: Data are normal
H1: Data are not normal
In this case, assuming an alpha risk of 0.05, we fail to reject our null
hypothesis and state our data are normal enough for what we want to
accomplish today! Had this P value been less than 0.05 we would have
rejected the null and stated the data do not appear normal enough for
Research Methodology For Management Decision
us. But no worries we are ready to press on with our stable and normal
data.
Setup the One Sample T-Test
We are now ready to run the one sample t-test. In this situation we will
be stating our null and alternate hypothesis a bit differently so please
stay focused. If you remember back to our original objective we are only
concerned if our critical dimension is less than 26 mm. In other words,
we need our critical dimension to be equal to or greater than 26 mm. So
we set things up as follows.
Ho: mu > or = 26
H1: mu < 26
To accomplish this in Minitab we have forced the alternate hypothesis to
be less than by clicking the options button in the one sample t-test
window which then brings us to this window where we make the change.
References