Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Yes, research is about finding facts, analyzing them and then evaluating the results.
Research begins with a theory or thesis statement. This statement has to be proven or
disproven as true/false. Given that research is based on proving or disproving a theory,
it means that research is concerned with finding facts, analyzing facts and evaluating
the response, in order to form a conclusion regarding the research topic.
There have been occasions, when research has been called into question. Corporate
espionage and other issues such as; corporations skewing the facts to fit their data have
existed. It is a downside to dishonesty in people; however, the ethical, moral, and
scientific belief is that all research has to be based on facts.
This does not mean the facts will not change. Research is conducted with what
information and tools are available at that time. This means that in 100 years research
being conducted now could be found false, but at the time it is true because of the
limited technology or facts that could be found.
As always when someone learns about research and the research method, one
is told that a theory is never solely factual, but proved or disproved based on what could
be found at that time. It goes back to the fact, that proper information, analysis, and
evaluation are needed in order to conductproper research. Inaccurate facts will skew the
data and render the entire research invalid.
There is also the human interpretation of the information found. While research is
concerned with these three topics, one also has to realize the writer of the research can
limit the scope of the research and thereforechange the results based on
their viewpoint alone.
Many times the initial problem identified in the first step of the process is too large or broad in scope. In step 3 of the
process, the researcher clarifies the problem and narrows the scope of the study. This can only be done after the
literature has been reviewed. The knowledge gained through the review of literature guides the researcher in
clarifying and narrowing the research project. In the example, the programmer has identified childhood obesity as
the problem and the purpose of the study. This topic is very broad and could be studied based on genetics, family
environment, diet, exercise, self-confidence, leisure activities, or health issues. All of these areas cannot be
investigated in a single study; therefore, the problem and purpose of the study must be more clearly defined. The
programmer has decided that the purpose of the study is to determine if walking 10,000 steps a day for three days a
week will improve the individuals health. This purpose is more narrowly focused and researchable than the original
problem.
Step 4: Clearly Define Terms and Concepts
Terms and concepts are words or phrases used in the purpose statement of the study or the description of the study.
These items need to be specifically defined as they apply to the study. Terms or concepts often have different
definitions depending on who is reading the study. To minimize confusion about what the terms and phrases mean,
the researcher must specifically define them for the study. In the obesity study, the concept of individuals health
can be defined in hundreds of ways, such as physical, mental, emotional, or spiritual health. For this study, the
individuals health is defined as physical health. The concept of physical health may also be defined and measured
in many ways. In this case, the programmer decides to more narrowly define individual health to refer to the areas
of weight, percentage of body fat, and cholesterol. By defining the terms or concepts more narrowly, the scope of the
study is more manageable for the programmer, making it easier to collect the necessary data for the study. This also
makes the concepts more understandable to the reader.
Step 5: Define the Population
Research projects can focus on a specific group of people, facilities, park development, employee evaluations,
programs, financial status, marketing efforts, or the integration of technology into the operations. For example, if a
researcher wants to examine a specific group of people in the community, the study could examine a specific age
group, males or females, people living in a specific geographic area, or a specific ethnic group. Literally thousands of
options are available to the researcher to specifically identify the group to study. The research problem and the
purpose of the study assist the researcher in identifying the group to involve in the study. In research terms, the
group to involve in the study is always called the population. Defining the population assists the researcher in
several ways. First, it narrows the scope of the study from a very large population to one that is manageable.
Second, the population identifies the group that the researchers efforts will be focused on within the study. This
helps ensure that the researcher stays on the right path during the study. Finally, by defining the population, the
researcher identifies the group that the results will apply to at the conclusion of the study. In the example in table 2.4,
the programmer has identified the population of the study as children ages 10 to 12 years. This narrower population
makes the study more manageable in terms of time and resources.
Step 6: Develop the Instrumentation Plan
The plan for the study is referred to as the instrumentation plan. The instrumentation plan serves as the road map for
the entire study, specifying who will participate in the study; how, when, and where data will be collected; and the
content of the program. This plan is composed of numerous decisions and considerations that are addressed in
chapter 8 of this text. In the obesity study, the researcher has decided to have the children participate in a walking
program for six months. The group of participants is called the sample, which is a smaller group selected from the
population specified for the study. The study cannot possibly include every 10- to 12-year-old child in the community,
so a smaller group is used to represent the population. The researcher develops the plan for the walking program,
indicating what data will be collected, when and how the data will be collected, who will collect the data, and how the
data will be analyzed. The instrumentation plan specifies all the steps that must be completed for the study. This
ensures that the programmer has carefully thought through all these decisions and that she provides a step-by-step
plan to be followed in the study.
Step 7: Collect Data
Once the instrumentation plan is completed, the actual study begins with the collection of data. The collection of data
is a critical step in providing the information needed to answer the research question. Every study includes the
collection of some type of datawhether it is from the literature or from subjectsto answer the research question.
Data can be collected in the form of words on a survey, with a questionnaire, through observations, or from the
literature. In the obesity study, the programmers will be collecting data on the defined variables: weight, percentage
of body fat, cholesterol levels, and the number of days the person walked a total of 10,000 steps during the class.
The researcher collects these data at the first session and at the last session of the program. These two sets of data
are necessary to determine the effect of the walking program on weight, body fat, and cholesterol level. Once the
data are collected on the variables, the researcher is ready to move to the final step of the process, which is the data
analysis.
Step 8: Analyze the Data
All the time, effort, and resources dedicated to steps 1 through 7 of the research process culminate in this final step.
The researcher finally has data to analyze so that the research question can be answered. In the instrumentation
plan, the researcher specified how the data will be analyzed. The researcher now analyzes the data according to the
plan. The results of this analysis are then reviewed and summarized in a manner directly related to the research
questions. In the obesity study, the researcher compares the measurements of weight, percentage of body fat, and
cholesterol that were taken at the first meeting of the subjects to the measurements of the same variables at the final
program session. These two sets of data will be analyzed to determine if there was a difference between the first
measurement and the second measurement for each individual in the program. Then, the data will be analyzed to
determine if the differences are statistically significant. If the differences are statistically significant, the study
validates the theory that was the focus of the study. The results of the study also provide valuable information about
one strategy to combat childhood obesity in the community.
As you have probably concluded, conducting studies using the eight steps of the scientific research process requires
you to dedicate time and effort to the planning process. You cannot conduct a study using the scientific research
process when time is limited or the study is done at the last minute. Researchers who do this conduct studies that
result in either false conclusions or conclusions that are not of any value to the organization.
This is an excerpt from Applied Research and Evaluation Methods in Recreation.
Research objective(s)
What are the research objectives?
In general, research objectives describe what we expect to achieve by a
project.
Research objectives are usually expressed in lay terms and are directed as
much to the client as to the researcher. Research objectives may be linked
with a hypothesis or used as a statement of purpose in a study that does not
have a hypothesis.
Even if the nature of the research has not been clear to the layperson from the
hypotheses, s/he should be able to understand the research from the
objectives.
A statement of research objectives can serve to guide the activities of
research. Consider the following examples.
Objective: To describe what factors farmers take into account in
making such decisions as whether to adopt a new technology or
what crops to grow.
Objective: To develop a budget for reducing pollution by a
particular enterprise.
Objective: To describe the habitat of the giant panda in China.
In the above examples the intent of the research is largely descriptive.
In the case of the first example, the research will end the study by
being able to specify factors which emerged in household decisions.
In the second, the result will be the specification of a pollution
reduction budget.
Although the research works and studies differ in their form and kind, they all still meet on the common
ground of scientific methods employed by them. Hence, scientific research is expected to satisfy the
following criteria:
i. The aim of the research should be clearly mentioned, along with the use of common concepts.
ii. The procedures used in the research should be adequately described, in order to permit another
researcher to repeat the research for further advancement, while maintaining the continuity of what has
already been done.
iii. The researchs procedural design should be carefully planned to obtain results that are as objective as
possible.
iv. The flaws in the procedural design should be sincerely reported by the researcher to correctly
estimate their effects upon the findings.
v. The data analysis should be adequate to reveal its significance.
vi. The methods used during the analysis should be appropriate.
vii. The reliability and validity of the concerned data should be checked carefully.
viii. The conclusions are needed to be confined and limited to only those data, which are justified and
adequately provided by the research.
ix. In case, the researcher is experienced and has a good reputation in the field of research, greater
confidence in research is warranted.
BY Ashish Soni
IN Research Methodology
WITH No Comments
Every research study, irrespective of its type, should meet some criterions so that it can be classified
as good research. Let us discuss some of these criterions.
1.
One of the important characteristics of a good research is that the purpose of the research is
clearly defined. A research study with clearly defined purpose finds a wider acceptance and
acknowledgement within the research community.
2.
Second important characteristic of a good research is that the research method should be
defined in a clear manner with sufficient detail. This will allow the repetition of the study in future
for further advancement, while maintaining the continuity of what has been done in the past.
3.
The third thing to remember is that any limitations and assumptions made by the researcher
during the course of the study should be clearly highlighted in the research. This will support the
findings of the research study, in case someone tries to validate the study findings.
4.
The fourth thing to remember is that, as far as possible, the research design should be planned
in a way that the results generated are as objective as possible. This will provide an easier
understanding about the findings of the research.
5.
Another thing to be considered by the researcher is that there should be sufficient data to
investigate the research topic. And the researcher should carefully check the reliability and
validity of the data.
6.
Further, in order to deliver a good research, a researcher should confine the conclusions to
those justified by the data.
7.
Lastly, a good research depends a great deal on the integrity and commitment of the
researcher.
Lack of Scientific Training: The research methodology is not systematic. Many researchers undertake
research work without having actual knowledge of the research methods. They just look for similar studies
and copy the methodologies listed in it. Even the research guides do not have a thorough knowledge of
the various methodologies. This scenario warrants the need for some sort of short-term training to be
imparted to researchers prior to undertaking research activities.
Insufficient Interaction: There is no proper interaction between researchers and the business
establishments, government institutions, etc. This leads to a great deal of data going untapped.
Interaction programs should be organized between researchers and other institutions on a regular basis.
This will highlight what issues need to be researched, what data is required for conducting research, and
how the study will be useful.
Lack of Confidence: Most of the business establishments are of the opinion that, researchers can
misuse the data provided by them. As such, they are reluctant to divulge details of their company. This
affects the research studies for which that particular data may be of utmost importance. Thus, confidencebuilding measures should be adopted, which will convince the business units that their data will be put to
productive purposes, and will not be misused in any manner by the researcher.
Lack of Code of Conduct: No specific code of conduct exists for the researchers, which leads to interdepartmental and inter-university rivalries.
Inadequate Assistance: Researchers in India have to cope with the non-availability of adequate and
timely secretarial assistance, which affects the schedule of their research study.
Improper Library Management: The libraries are not managed systematically. Much of the precious time
of the researchers is spent in looking for books, reports, newspapers, etc. rather than searching relevant
information from them.
High Cost of Publishing: Once their research is completed, the researchers have to look for a means to
publish it. Publishing in international journals is highly expensive. This discourages most of the
researchers from taking up research work.
drug, the null hypothesis might be that the new drug is no better, on average,
than the current drug. We would write
H0: there is no difference between the two drugs on average.
We give special consideration to the null hypothesis. This is due to the fact
that the null hypothesis relates to the statement being tested, whereas the
alternative hypothesis relates to the statement to be accepted if / when the null
is rejected.
The final conclusion once the test has been carried out is always given in
terms of the null hypothesis. We either "Reject H0 in favour of H1" or "Do not
reject H0"; we neverconclude "Reject H1", or even "Accept H1".
If we conclude "Do not reject H0", this does not necessarily mean that the null
hypothesis is true, it only suggests that there is not sufficient evidence against
H0 in favour of H1. Rejecting the null hypothesis then, suggests that the
alternative hypothesis may be true.
See also hypothesis test.
Alternative Hypothesis
The alternative hypothesis, H1, is a statement of what a statistical hypothesis
test is set up to establish. For example, in a clinical trial of a new drug, the
alternative hypothesis might be that the new drug has a different effect, on
average, compared to that of the current drug. We would write
H1: the two drugs have different effects, on average.
The alternative hypothesis might also be that the new drug is better, on
average, than the current drug. In this case we would write
H1: the new drug is better than the current drug, on average.
The final conclusion once the test has been carried out is always given in
terms of the null hypothesis. We either "Reject H0 in favour of H1" or "Do not
reject H0". We never conclude "Reject H1", or even "Accept H1".
If we conclude "Do not reject H0", this does not necessarily mean that the null
hypothesis is true, it only suggests that there is not sufficient evidence against
H0 in favour of H1. Rejecting the null hypothesis then, suggests that the
alternative hypothesis may be true.
Alternative Hypothesis
An alternative hypothesis states that there is statistical significance between two variables. In the
earlier example, the two variables are Mentos and Diet Coke. The alternative hypothesis is the
hypothesis that the researcher is trying to prove. In the Mentos and Diet Coke experiment, Arnold
was trying to prove that the Diet Coke would explode if he put Mentos in the bottle. Therefore, he
proved his alternative hypothesis was correct.
If we continue with example, the alternative hypothesis would be that there IS indeed a statistically
significant relationship between Mentos and Diet Coke. Arnold could write it as:
If I put half a pack of Mentos into a 2-Liter Diet Coke bottle, there will be a big reaction/explosion.
If we continue with the example, the null hypothesis Arnold would write is: if I put half a pack of
Mentos into a 2-Liter Diet Coke bottle, there will be no reaction or explosion.
2.Stratified sampling
From Wikipedia, the free encyclopedia
3.Random Sampling
ssimple Random Sampling
Simple random sampling is the most widely-used probability sampling method, probably
because it is easy to implement and easy to analyze.
Key Definitions
To understand simple random sampling, you need to first understand a few key definitions.
The total set of observations that can be made is called the population.
There are many ways to obtain a simple random sample. One way would be the lottery
method. Each of the N population members is assigned a unique number. The numbers are
placed in a bowl and thoroughly mixed. Then, a blind-folded researcher selects n numbers.
Population members having the selected numbers are included in the sample.
replacement is approximately the same as sampling with replacement, since the odds of choosing
the same individual twice is low.
It is critical because it relates to the types of statistics you can use to analyse your
data. An easy way to have a paper rejected is to have used either an incorrect
scale/statistic combination or to have used a low powered statistic on a high powered
set of data.
Nominal
Ordinal
Interval
Ratio
Nominal
The lowest measurement level you can use, from a statistical point
of view, is a nominal scale.
A nominal scale, as the name implies, is simply some placing of data into
categories, without any order or structure.
A physical example of a nominal scale is the terms we use for colours. The
underlying spectrum is ordered but the names are nominal.
In research activities a YES/NO scale is nominal. It has no order and there is
no distance between YES and NO.
and statistics
The statistics which can be used with nominal scales are in the nonparametric group. The most likely ones would be:
mode
crosstabulation - with chi-square
There are also highly sophisticated modelling techniques
available for nominal data.
Ordinal
An ordinal scale is next up the list in terms of power of
measurement.
The simplest ordinal scale is a ranking. When a market researcher asks you to
rank 5 types of beer from most flavourful to least flavourful, he/she is asking
you to create an ordinal scale of preference.
There is no objective distance between any two points on your subjective
scale. For you the top beer may be far superior to the second prefered beer
but, to another respondant with the same top and second beer, the distance
may be subjectively small.
An ordinal scale only lets you interpret gross order and not the relative
positional distances.
and statistics
Interval
When you are asked to rate your satisfaction with a piece of software on a 7
point scale, from Dissatisfied to Satisfied, you are using an interval scale.
It is an interval scale because it is assumed to have equidistant points between
each of the scale elements. This means that we can interpret differences in the
distance along the scale. We contrast this to an ordinal scale where we can
only talk about differences in order, not differences in the degree of order.
Interval scales are also scales which are defined by metrics such as
logarithms. In these cases, the distances are note equal but they are strictly
definable based on the metric used.
and statistics
Ratio
A ratio scale is the top level of measurement and is not often
available in social research.
The factor which clearly defines a ratio scale is that it has a true zero point.
The simplest example of a ratio scale is the measurement of length
(disregarding any philosophical points about defining how we can identify
zero length).
The best way to contrast interval and ratio scales is to look at temperature.
The Centigrade scale has a zero point but it is an arbitrary one. The Farenheit
scale has its equivalent point at -32o. (Physicists would probably argue that
Absolute Zero is the zero point for temperature but this is a theoretical
concept.) So, even though temperture looks as if it would be a ratio scale it is
an interval scale. Currently, we cannot talk about no temperature - and this
would be needed if it were a ration scale.
and statistics
There are many benefits of data analysis however; the most important ones are as follows: - data
analysis helps in structuring the findings from different sources of data collection like survey
research. It is again very helpful in breaking a macro problem into micro parts. Data analysis
acts like a filter when it comes to acquiring meaningful insights out of huge data-set. Every
researcher has sort out huge pile of data that he/she has collected, before reaching to a
conclusion of the research question. Mere data collection is of no use to the researcher. Data
analysis proves to be crucial in this process. It provides a meaningful base to critical decisions. It
helps to create a complete dissertation proposal.
One of the most important uses of data analysis is that it helps in keeping human bias away from
research conclusion with the help of proper statistical treatment. With the help of data analysis a
researcher can filter both qualitative and quantitative data for an assignment writing projects.
Thus, it can be said that data analysis is of utmost importance for both the research and the
researcher. Or to put it in another words data analysis is as important to a researcher as it is
important for a doctor to diagnose the problem of the patient before giving him any treatment.
The main portion of Statistics is the display of summarized data. Data is initially collected from a
given source, whether they are experiments, surveys, or observation, and is presented in one of four
methods:
Textular Method
(Don`t base with this info, it`s hoax!) the gathered data.
Tabular Method
Provides a more precise, systematic and orderly presentation of data in rows or columns.
Semi-tabular Method
Advertisements:
The activities of an organisation rest upon collection of factual data and information. The statistical
investigator collects the primary data from the original sources and secondary data are collected
from the published sources. Various methods of collection of primary data are:
(a) Direct personal interview-Under this method, data are collected by the investigator through
direct conversation with the people.
(b) Indirect investigation under this method, the investigator collects information indirectly by the
contact of a third party.
(c) Observation and experiment under this method, information is collected by conducting
observation and experiment.
(d) Correspondents under this method, correspondents are appointed in different localities to
collect information.
(e) Questionnaire under this method, the information is collected by sending a set of questions in
questionnaire.
(f) Secondary data are collected from published sources like magazines and periodicals.
A one- or two-tailed t-test is determined by whether the total area of is placed in one
tail or divided equally between the two tails. The one-tailed t-test is performed if the
results are interesting only if they turn out in a particular direction. The two-tailed ttest is performed if the results would be interesting in either direction. The choice of a
one- or two-tailed t-test effects the hypothesis testing procedure in a number of
different ways.
TWO-TAILED t-TESTS
A two-tailed t-test divides in half, placing half in the each tail. The null hypothesis
in this case is a particular value, and there are two alternative hypotheses, one positive
and one negative. The critical value of t, t crit, is written with both a plus and minus sign
( ). For example, the critical value of t when there are ten degrees of freedom (df=10)
and is set to .05, is tcrit= 2.228. The sampling distribution model used in a twotailed t-test is illustrated below:
ONE-TAILED t-TESTS
There are really two different one-tailed t-tests, one for each tail. In a one-tailed t-test,
all the area associated with is placed in either one tail or the other. Selection of the
tail depends upon which direction tobs would be (+ or -) if the results of the experiment
came out as expected. The selection of the tail must be made before the experiment is
conducted and analyzed.
A one-tailed t-test in the positive direction is illustrated below:
The value tcrit would be positive. For example when is set to .05 with ten degrees of
freedom (df=10), tcrit would be equal to +1.812.
A one-tailed t-test in the negative direction is illustrated below:
The value tcrit would be negative. For example, when is set to .05 with ten degrees of
freedom (df=10), tcrit would be equal to -1.812.
One-tailed tests
In the previous pages, you learned how to perform define the hypothesis for a
statistical test, then to perform a t-test to compare means. In the example t-test we
performed, we defined an alternate hypothesis to test whether one mean was greater
than the other: > 0.
In this situation, we tested whether one mean was higher than the other. We were not
interested in whether the first mean was lower than the other, only if it was higher. So
we were only interested in one side of the probability distribution, which is shown in
the image below:
In this distribution, the shaded region shows the area represented by the null
hypothesis, H0: = 0. This actually implies 0, since the only unshaded region in
the image shows > 0. Because we were only interested in one side of the
distribution, or one "tail", this type of test is called a one-sided or a one-tailed test.
When you are using tables for probability distributions, you should make sure whether
they are for one-tailed or two-tailed tests. Depending on which they are for, you need
to know how to switch to the one you need. This is all explained below.
A one-tailed test uses an alternate hypothesis that states
either H1: > 0 OR H1: < 0, but not both. If you want to test both, using the
alternate hypothesis H1: 0, then you need to use a two-tailed test.
Two-tailed tests
We would use a two-tailed test to see if two means are different from each other (ie
from different populations), or from the same population. As an example, let's assume
that we want to check if the pH of a stream has changed significantly in the past year.
A water sample from the stream was analyzed using a pH electrode, where six
samples were taken. It was found that the mean pH reading was 6.5 with standard
deviation sold = 0.2. A year later, six more samples were analyzed, and the mean pH of
these readings was 6.8 with standard deviation sold = 0.1.
Example 1
We could use a one-tailed test, to see if the stream has a higher pH than one year ago,
for which we would use the alternate hypothesis HA: prev < current. However, we may
want a more rigorous test, for the hypothesis that HA: prev current. This would mean
that both HA: prev < current and HA: prev > current were satisfied, and we could be sure
that there is a significant difference between the means. The probability distribution
for a 90% confidence level, two-tailed test looks like this:
Continuing the example, we define the null hypothesis H0: prev = current, and the
alternate hypothesis HA: prev current. The d.o.f. for a two sample mean t-test is =
7.35 7, since the d.o.f. must be a whole number. The t-value for the two sample test
is
If we consult a two-tailed t-test table, for a 95% confidence limit, we find that t7,95% =
2.36. Since tcalc > t7,95%, we reject the null hypothesis, accept the alternate hypothesis
that prev current, and can say that the means are significantly different.
Type I error
When the null hypothesis is true and you reject it, you make a type I error. The
probability of making a type I error is , which is the level of significance you set
for your hypothesis test. An of 0.05 indicates that you are willing to accept a
5% chance that you are wrong when you reject the null hypothesis. To lower this
risk, you must use a lower value for . However, using a lower value for alpha
means that you will be less likely to detect a true difference if one really exists.
Type II error
When the null hypothesis is false and you fail to reject it, you make a type II
error. The probability of making a type II error is , which depends on the power
of the test. You can decrease your risk of committing a type II error by ensuring
your test has enough power. You can do this by ensuring your sample size is large
There are 4 typical levels of confidence: 99%, 98%, 95% and 90%. Each of the
levels of confidence has a different number of standard errors associated
with it. We denote this by
where is the total amount of area in the tails of the normal curve. Thus,
for a 95% level of confidence
Significance area
(o-e)2/e
) is:
That is, chi-square is the sum of the squared difference between observed (o) and the
expected (e) data (or the deviation, d), divided by the expected data in all possible
categories.
For example, suppose that a cross between two pea plants yields a population of 880
plants, 639 with green seeds and 241 with yellow seeds. You are asked to propose the
genotypes of the parents. Yourhypothesis is that the allele for green is dominant to the
allele for yellow and that the parent plants were both heterozygous for this trait. If
your hypothesis is true, then the predicted ratio of offspring from this cross would be
3:1 (based on Mendel's laws) as predicted from the results of the Punnett square
(Figure B. 1).
A chi-squared test, also referred to as
test in which the sampling distribution of the test statistic is a chi-square distribution when the null
hypothesis is true. Chi-squared tests are often constructed from a sum of squared errors, or through
the sample variance. Test statistics that follow a chi-squared distribution arise from an assumption of
independent normally distributed data, which is valid in many cases due to the central limit theorem.
A chi-squared test can then be used to reject the null hypothesis that the data are independent.
Also considered a chi-square test is a test in which this is asymptotically true, meaning that the
sampling distribution (if the null hypothesis is true) can be made to approximate a chi-square
distribution as closely as desired by making the sample size large enough. The chi-squared test is
used to determine whether there is a significant difference between the expected frequencies and
the observed frequencies in one or more categories. Does the number of individuals or objects that
fall in each category differ significantly from the number you would expect? Is this difference
between the expected and observed due to sampling variation, or is it a real difference?
perform a one-way analysis of variance with stacked data 1 Choose Stat ANOVA
One-way. 2 In Response, enter the column containing the responses. 3 In Factor,
enter the column containing the factor levels. 4 If you like, use one or more of the
options described below, then click OK. STANOVAR.MK5 Page 5 Friday, December
17, 1999 12:33 PM Chapter 3 One-Way Analysis of Variance 3-6 MINITAB Users
Guide 2 CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE
CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE h To
perform a one-way analysis of variance with unstacked data 1 Choose Stat ANOVA
One-way (Unstacked). 2 In Responses (in separate columns), enter the columns
containing the separate response variables. 3 If you like, use one or more of the
options described below, then click OK. Options with stacked data One-way dialog
box store residuals and fitted values (the means for each level). Comparisons
subdialog box display confidence intervals for the differences between means,
using four different multiple comparison methods: Fishers LSD, Tukeys, Dunnetts,
and Hsus MCB (multiple comparisons with the best). See Multiple comparisons of
means on page 3-7. Graphs subdialog box draw boxplots, dotplots, and residual
plots. You can draw five different residual plots: histogram. normal probability
plot. plot of residuals versus the fitted values ( ). plot of residuals versus data
order. The row number for each data point is shown on the x-axisfor example, 1 2
3 4 n. separate plot for the residuals versus each specified column. For a
discussion of the residual plots, see Residual plots on page 2-6. Y STANOVAR.MK5
Page 6 Friday, December 17, 1999 12:33 PM One-Way Analysis of Variance Analysis
of Variance MINITAB Users Guide 2 3-7 CONTENTS INDEX MEET MTB UGUIDE 1
UGUIDE 2 SC QREF HOW TO USE CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2
SC QREF HOW TO USE Options with unstacked data Graphs subdialog box draw
boxplots and dotplots that display the sample mean for each sample. Multiple
comparisons of means Multiple comparisons of means allow you to examine which
means are different and to estimate by how much they are different. When you
have a single factor and your data are stacked, you can obtain multiple comparisons
of means by choosing the Stat ANOVA One-way and then clicking the
Comparisons subdialog box. The choice of method The multiple comparison
methods compare different means and use different error rates. If you wish to
examine all pairwise comparisons of means, use either Fishers least significant
difference (LSD) or Tukeys (also called Tukey-Kramer in the unbalanced case)
method. The choice depends on whether you wish to control the individual
(comparison-wise) error rate or the family (experiment-wise) error rate. The danger
in using the individual error rate with Fishers method is having an unexpectedly
high probability of making at least one Type I error (declaring a difference when
there is none) among all the comparisons. MINITAB displays both error rates. In
most cases, the Tukey method is probably the choice that you should make when
you want to judge all pairwise differences, because you can control the family error
rate. Choose the Dunnett method if you are comparing treatments to a control.
When this method is suitable, it is inefficient to use the Tukey all-pairwise approach,
because the Tukey confidence intervals will be wider and the hypothesis tests less
powerful for a given family error rate. You will need to specify which level represents
the control. If this level is text or date/time, enclose it with double quotes. Choose
Hsus MCB (multiple comparison with the best) method if it makes sense to compare
each mean only with the best among all of the other ones. This procedure allows
you to judge how much worse a level might be if it is not the best or how much
better it might be than its closest competitor. You will need to specify if the best is
smallest or largest. If you are mainly interested in comparing each level to the
best it is inefficient to use the Tukey all-pairwise approach because you will waste
your error rate comparing pairs of level means which do not include the best mean.
STANOVAR.MK5 Page 7 Friday, December 17, 1999 12:33 PM Chapter 3 One-Way
Analysis of Variance 3-8 MINITAB Users Guide 2 CONTENTS INDEX MEET MTB
UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE CONTENTS INDEX MEET MTB UGUIDE 1
UGUIDE 2 SC QREF HOW TO USE Your choice among them may depend on their
properties. Some properties of the multiple comparison methods are summarized
below: Interpreting confidence intervals MINITAB presents results in confidence
interval form to allow you to assess the practical significance of differences among
means, in addition to statistical significance. As usual, the null hypothesis of no
difference between means is rejected if and only if zero is not contained in the
confidence interval. Specify error rates as percents between 0.1 and 50%. The
default error rate of 5% is the family error rate for the Tukey, Dunnett, and MCB
methods and the individual error rate for the Fisher method. Individual error rates
are exact in all cases, meaning that they can be calculated by an explicit formula.
Family error rates are exact for equal group sizes. If group sizes are unequal, the
true family error rate for the Tukey, Fisher, and MCB methods will be slightly smaller
than stated, resulting in conservative confidence intervals [4], [21]. The Dunnett
family error rates are exact for unequal sample sizes. The F-test and multiple
comparisons The results of the F-test and multiple comparisons can conflict. For
example, it is possible for the F-test to reject the null hypothesis of no differences
among the level means, and yet all the Tukey pairwise confidence intervals may
contain zero. Conversely, it is possible for the F-test to fail to reject the null
hypothesis, and yet have one or more of the Tukey pairwise confidence intervals not
include zero. The F-test has been used to protect against the occurrence of false
positive differences in means. However, the Tukey, Dunnett, and MCB methods have
protection against false positives built in, while the Fisher method only benefits from
this protection when all means are equal. If the use of multiple comparisons is
conditioned upon the significance of the F-test, the error rate can be higher than the
error rate in the unconditioned application of multiple comparisons [14]. See Help
for computational details of the multiple comparison methods. Comparison method
Purpose Error rate Fishers LSD all pairwise differences individual Tukey all pairwise
differences family Dunnett comparison to a control family Hsus MCB comparison
with the best family STANOVAR.MK5 Page 8 Friday, December 17, 1999 12:33 PM
One-Way Analysis of Variance Analysis of Variance MINITAB Users Guide 2 3-9
CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE CONTENTS
INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE e Example of a one-
way ANOVA with multiple comparisons You design an experiment to assess the
durability of four experimental carpet products. You place a sample of each of the
carpet products in four homes and you measure durability after 60 days. Because
you wish to test the equality of means and to assess the differences in means, you
use the one-way ANOVA procedure (data in stacked form) with multiple
comparisons. Generally, you would choose one multiple comparison method as
appropriate for your data. However, two methods are selected here to demonstrate
MINITABs capabilities. 1 Open the worksheet EXH_AOV.MTW. 2 Choose Stat
ANOVA One-way. 3 In Response, enter Durability. In Factor, enter Carpet. 4 Click
Comparisons. Check Tukeys, family error rate and enter 10 in the text box. Check
Hsus MCB, family error rate and enter 10 in the text box. Click OK in each dialog
box. Session window output One-way ANOVA: Durability versus Carpet Analysis of
Variance for Durabili Source DF SS MS F P Carpet 3 111.6 37.2 2.60 0.101 Error 12
172.0 14.3 Total 15 283.6 Individual 95% CIs For Mean Based on Pooled StDev Level
N Mean StDev ---------+---------+---------+------- 1 4 14.483 3.157 (-------*-------) 2 4
9.735 3.566 (-------*--------) 3 4 12.808 1.506 (--------*-------) 4 4 17.005 5.691 (------*-------) ---------+---------+---------+------- Pooled StDev = 3.786 10.0 15.0 20.0 Hsus
MCB (Multiple Comparisons with the Best) Family error rate = 0.100 Critical value =
1.87 Intervals for level mean minus largest of other level means Level Lower Center
Upper -+---------+---------+---------+------ 1 -7.527 -2.522 2.482 (--------*-------) 2 -12.274
-7.270 0.000 (-------*-----------) 3 -9.202 -4.198 0.807 (-------*-------) 4 -2.482 2.522
7.527 (-------*--------) -+---------+---------+---------+------ -12.0 -6.0 0.0 6.0
STANOVAR.MK5 Page 9 Friday, December 17, 1999 12:33 PM Chapter 3 One-Way
Analysis of Variance 3-10 MINITAB Users Guide 2 CONTENTS INDEX MEET MTB
UGUIDE 1 UGUIDE 2 SC QREF HOW TO USE CONTENTS INDEX MEET MTB UGUIDE 1
UGUIDE 2 SC QREF HOW TO USE Tukeys pairwise comparisons Family error rate =
0.100 Individual error rate = 0.0250 Critical value = 3.62 Intervals for (column level
mean) - (row level mean) 1 2 3 2 -2.106 11.601 3 -5.178 -9.926 8.528 3.781 4
-9.376 -14.123 -11.051 4.331 -0.417 2.656 Interpreting the results The default oneway output contains an analysis of variance table, a table of level means, individual
95% confidence intervals, and the pooled standard deviation. The F-test p-value of
0.101 indicates that there is not quite sufficient evidence (at = 0.10 or less) to
claim that not all the means are equal. However, you should examine the multiple
comparison results, which use family error rates of 0.10, because the methods used
(Tukey, MCB) have built in protection against false positive results. The output
labeled Hsus MCB compares each mean with the best of the other means. Here,
best is the default or largest of the others. The means of carpets 1, 2, and 3 were
compared to the level 4 mean because the carpet 4 mean is the largest of the rest.
The level 4 mean was compared to the carpet 1 mean. Carpets 1, 3, or 4 may be
best, since the corresponding confidence intervals contain positive values. There is
no evidence that carpet 2 is the best because the upper interval endpoint is 0, the
smallest it can be. In addition, it is possible to describe the potential advantage or
disadvantage of any of the contenders for the best. For example, if carpet 3 is best,
it is no more than 0.809 better than its closest competitor, and it may be as much
as 9.204 worse than the best of the other level means. If carpet 1 is not the best, it
is no more than 2.484 worse than the best of the other means, and it may be as
much as 7.529 better than the best of the others. The first pair of numbers in the
Tukey output table, (2.106, 11.601), gives the confidence interval for the mean of
carpet 1 minus the mean of carpet 2. Confidence intervals for entries not in the
table can be found from entries in the table. For example, the confidence interval
for the mean of level 2 minus the mean of carpet 1 is (11.601, 2.106). Carpets 2
and 4 are the only ones for which the means can be declared as different, since the
confidence interval for this combination of means is the only one that excludes zero.
STANOVAR.MK5 Page 10 Friday, December 17, 1999 12:33 PM
s
and 95% confidence intervals for factor levels means. store residuals and fits. fit
an additive model, that is, a model without the interaction term. In this case, the
fitted value for cell (i, j) is (mean of observations in row i) + (mean of observations
in row j) (mean of all observations). Graphs subdialog box draw five different
residual plots. You can display the following plots: histogram. normal probability
plot. plot of residuals versus the fitted values ( ). plot of residuals versus data
order. The row number for each data point is shown on the x-axisfor example, 1 2
3 4 n. separate plot for the residuals versus each specified column. For a
discussion of the residual plots, see Residual plots on page 2-6. e Example of twoway analysis of variance You are a biologist who is studying how zooplankton live in
two lakes. You set up twelve tanks in your laboratory, six each with water from a
different lake. You add one of three nutrient supplements to each tank and after 30
days you count the zooplankton in a unit volume of water. You use two-way ANOVA
to test if the population means are equal, or equivalently, to test whether there is
significant evidence of interactions and main effects. 1 Open the worksheet
EXH_AOV.MTW. 2 Choose Stat ANOVA Two-way. 3 In Response, enter
Zooplankton. 4 In Row factor, enter Supplement. Check Display means. 5 In Column
factor, enter Lake. Check Display means. Click OK. Y STANOVAR.MK5 Page 12
Friday, December 17, 1999 12:33 PM Analysis of Means Analysis of Variance
MINITAB Users Guide 2 3-13 CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC
QREF HOW TO USE CONTENTS INDEX MEET MTB UGUIDE 1 UGUIDE 2 SC QREF HOW
TO USE Session window output Two-way ANOVA: Zooplankton versus Supplement,
Lake Analysis of Variance for Zooplank Source DF SS MS F P Suppleme 2 1919 959
9.25 0.015 Lake 1 21 21 0.21 0.666 Interaction 2 561 281 2.71 0.145 Error 6 622
104 Total 11 3123 Individual 95% CI Suppleme Mean --+---------+---------+--------+--------- 1 43.5 (-------*-------) 2 68.3 (--------*-------) 3 39.8 (--------*-------) --+--------+---------+---------+--------- 30.0 45.0 60.0 75.0 Individual 95% CI Lake Mean -----+---------+---------+---------+----- Dennison 51.8 (----------------*----------------) Rose 49.2
(----------------*----------------) ------+---------+---------+---------+----- 42.0 48.0 54.0 60.0
Interpreting the results The default output for two-way ANOVA is the analysis of
variance table. For the zooplankton data, there is no significant evidence for a
supplementlake interaction effect or a lake main effect if your acceptable value
is less than 0.145 (the p-value for the interaction F-test). There is significant
evidence for supplement main effects, as the F-test p-value is 0.015. As requested,
the means are displayed with individual 95% confidence intervals. Supplement 2
appears to have provided superior plankton growth in this experiment. These are tdistribution confidence intervals calculated using the error degrees of freedom and
the pooled standard deviation (square root of the mean square error). If you want to
examine simultaneous differences among means using multiple comparisons, use
General Linear Model