C.R.M.E. is a workbook for anyone needing (or wanting) to do a research project that can stand the test of claiming scientific significance. The workbook leads the reader in a logical progression from the measurement of variables in the research design
CREDIBLE RESEARCH MADE EASY
A Step by Step Path to Formulating Testable HypothesesBy DORCAS MLADENKAiUniverse, Inc.
Copyright © 2012 Dorcas Mladenka
All right reserved.ISBN: 978-1-4759-2030-7Contents
Acknowledgements....................................................................viiIntroduction........................................................................ixPart 1 - A Sample Project; Defining and Measuring Variables.........................1Part 2 - Formulating Hypotheses.....................................................10Part 3 - Collecting the Data........................................................21Part 4 - Theoretical Distributions and Looking for Significance.....................25Part 5 - Completing Your Own Research Project.......................................32Appendix: Outline and Notes for a Research Report...................................37
Chapter One
PART 1 A SAMPLE PROJECT; DEFINING AND MEASURING VARIABLES
Professor M teaches the business research course. She wants all of her students to do well in the course, and she wonders what elements might account for differences in their success levels. The problem area, then, is the research course and the success (or failure) of students taking the course.
Professor M decides that "success" in the course (DEPENDENT VARIABLE—DV) would have to be determined by the final grade in the course. She also decides, upon reflection, that since all students get the same material and the same assistance in the course, there may be other elements (INDEPENDENT VARIABLES—IVs) involved. She determines to explore these variables through an analytical, investigative research project.
You will use Professor M's research project as a sample to follow in doing your own research. This first section will include the following:
-> Research Design
-> Research Questions
-> Measurement of Variables
-> Operational Definitions
THE RESEARCH DESIGN
Let's assume the following research design. Prof. M is interested in what variables influence the grade that research students made in the business research course (your dependent variable). She selects five independent variables to include in her research, and a diagram of her research model looks like this.
Gender
Major Semester Hrs Research Grade Study for the Course Interest in Statistics (IVs) (DV)
PROBLEM STATEMENT: (Ask a question about the variables)To what extent do the variables of gender, major, course load, study, and interest in statistics predict the final research grade.
THE RESEARCH QUESTIONS (Another way of explaining your research Design)
You will now make the sample project your research as you study the process. You are theorizing relationships among your variables.
You are interested in the relationship that each of the IVs had to the dependent variable: how does each IV relate to the research grade? You will indicate the kind of relationship you are interested in researching by formulating questions—the research questions. In order to do that you will of course need a measurement of each variable, and that measurement has to be a quantity—a number.
RQ 1 So lets get started, first of all, what do you mean by the variable. When you say "research grade," you mean, of course, the average (mean) research grade made by students ... and you must specify, for example, students who took the course in the last two semesters: Spring XX or Fall XX.
Your first research question of interest then is a research grade—a very quantifiable measure. The mean of all the research grades in your sample is the measure for your DV. (The next section is on Measurement Scales and Operational Definitions)
RQ 2Was there a difference between male and female research grades? (IV Gender: DV Grade)
RQ 3 Did the research grade differ according to the students' majors? (IV Major: DV Grade)
RQ 4 Did the total number of semester hours a student carried during the semester the research was taken influence the research grade? (IV Semester Hrs.: DV Grade)
RQ 5 Did the amount of Study have an influence on the research grade? First you have to "clean up" ambiguous terms: amount ... study ... By "amount" do your mean how many hours or how often? By "study" do you mean specifically studying research? (IV Study: DV Grade)
RQ 6 Was there an association between how "interested in statistics" the student was and that student's research grade? Again you have to make clear what "interest" really means. (IV Interest in Statistics: DV Research grade.)
Good news: you are getting closer to developing hypotheses for your research project.
So far you have formulated six research questions: one having to do with data for the DV, the other five having to do with the relationship of each IV to the DV. In order to learn how to do a few more statistical tests, you will also look at relationships of one IV to another IV. There is, however, a more important reason for doing this. You will have collected the data, so it will be a simple matter to do the relationships. And you might find some surprises.
In this example we will add the following research questions:
RQ 7. Were marketing majors "more interested" in learning research than other majors? (IV 2: IV 5)
RQ 8 Was there a difference in study times between male and female students? (IV 1: IV 4)
There are two things you must do before your begin stating your research questions in the form of hypotheses.
1. You must be very precise about what you mean by the variable because you have to quantify it—make it measurable—one number for each variable.
2. You must determine how you will "tap" the variable—how you will collect the data.
The next two sections will help you perform these two tasks.
MEASUREMENT SCALES
Once you have determined the variables to use in your study, the big challenge is to "tap" them—to get the data on them. The data you get has to be in quantitative form—in numbers to feed into the computer.
NOTE: You won't be having to use formulas to work the tests—computers do that for you. A common software for this type of research is SPSS—available in schools and businesses. There are several measurement scales that are used in research. The most common are
nominal: the variable has no quantitative value as such; the variable is divided into groups, such as gender MALe / FeMALe (divides sample into categories)
ordinal: no quantitative value; ranks preferences
Rank the following cereals in order of your personal preference ______ cheerios ______ bran flakes ______ corn flakes
(differences; median)
interval: no quantitative value, number scale to show importance:
How much do you agree with the company's health insurance plan?
Strongly disagree Disagree Neutral Agree Strongly Agree 1 2 3 4 5
(can yield a mean on the scale)
ratio: meaningful measurement scale-absolute zero point
(research grade: ____)
(can yield a mean)
It is important to use the measurement scale that will give you the information you seek. Thus every variable has to be "operationally" defined.
OPERATIONAL DEFINITIONS
Every variable has to be operationally defined: for you so that you know what type of test to use ... and for the reader of your research. The reader of your research has to know precisely what you mean by the variable.
There are numerous types of statistical tests available to the researcher. In this workbook sample you will use a selected few. The tests you will use are commonly-used tests and cover a variety of possibilities for the way you might design your study. Learning how to use a few tests will also make it easier to learn additional tests as you need them.
The type of statistical test you use depends on the measurement scales that you used to tap your variables.
In order to know what statistical test to use and how to formulate your hypotheses, it is important to remember how you can tabulate and summarize using the various measurement scales.
NOMINAL: categories; groups; "samples" (categories cannot be averaged)
INTERVAL and RATIO: numbers (you can get a mean)
The way each variable was measured will dictate the type of relationship you will establish.
When you relate one variable to another, you look at how both variables were measured and that determines what kind of relationship you can make.
For example, you measured our DV, GRADES, using the ratio scale. You measured GENDER nominally. How do we relate GENDER to GRADES?
GENDER broke our sample into two groups. We can look at the "mean grade" for males and the "mean grade" for females and see the difference. The software will simply list all the male grades and get a mean, list all the female grades and get a mean, and show you the difference.
each nominal variable breaks your sample into two or more populations. When related to a ratio or interval scale, the relationship is a DIFFERENCE OF MEANS
Sometimes both variables are measured using a nominal scale: GENDER and STUDY.
Since you cannot get a "mean" from categories, you cannot apply a "means test," but you can use a test that will count the frequencies of the categories and determine whether the variables had anything to do with each other. THIS RELATIONSHIP IS A TEST OF "INDEPENDENCE."
SEMESTER HOURS and GRADES are both measured with a ratio scale. Neither measurement breaks the samples into groups. Instead, each gives us the "mean" for the whole sample.
The type of relationship to establish here is to see if there is a correlation-to see whether the two variable means move up at the same rate, or down at the same rate, or whether as one moves up the other moves down.
INTEREST is measured using the interval scale, which also yields a mean. So INTEREST and GRADES would also be tested for correlation.
The type of relationship established in both of the above pairs of variables is a TEST OF CORRELATION.
IMPORTANT: Each test issuited for a specific type of comparison. The variable measurements determine:
THE MEAN OF A POPULATION THE INDEPENDENCE OFTWO VARIABLES THE CORRELATION OFTWO VARIABLES A DIFFERENCE OF MEANS OF TWO POPULATIONS
TWO VARIABLE TESTS (BIVARIATE TESTS) compare the means of two or more populations
TESTING DIFFERENCES
Two Sample T Test compares the difference between the means of two populations
ANOVA compares the differences between the means of two or more populations
TESTING INDEPENDENCE
Chi Square tests for whether variables measured in nominal scale have an impact on one another
Two Sample
T Test compares the difference between the means of two populations
TESTING CORRELATION
Pearson's r
Correlation tests whether two variables are correlated
ONE VARIABLE TEST (UNIVARIATE TESTS)
One Sample
T Test predicts a mean
Research question No. 1 asked about the mean RESEARCH GRADE; in other words, your hypothesis will make a prediction about the variable. Only one variable is involved, so that requires a "one variable test" (above). All the other predictions will involve two variables and require "two variable tests" (above).
The table below gives a quick look at what test to use depending on how the variables are measured.
SHORTHAND FOR USING TESTS
If .... USE ...
V1 = NOMINAL (2) AND V2 = INTERVAL/RATIO T TEST FOR 2 MEANS V1 = NOMINAL (3+) AND V2 = INTERVAL/RATIO ANOVA V1 = NOMINAL AND V2 = NOMINAL CHI SQUARE V1 = INTERVAL/RATIO and V2 = INTERVAL/RATIO CORRELATION V1 = INTERVAL/RATIO T TEST FOR 1 MEAN
Although the hypotheses are formulated, the data collected and tabulated, and then the tests are applied, it is important to understand the relationship of measurement scales and statistical tests in order to state the hypothesis in an accurate, testable format.
The sections in PART 2 explain how to "state the hypothesis."
Chapter Two
PART 2 FORMULATING HYPOTHESES These are the steps in formulating your hypotheses:
1. A Research Design identifying the dependent variable and all independent variables to be included in the study.
2. Research questions: formulating research questions: what you are interested in finding out about the variables from your research.
3. Operational definitions: defining each variable in terms of a) what elements or aspects it encompasses; b) how it will be quantified.
4. Measuring the variables:
a) determining the measurement scale that will be most appropriate to "tap" the variable
b) formulating the question for the questionnaire.
5. Hypotheses statements: determining the type of hypothesis to use for each of the research questions (based on the measurement scale used) and stating the hypothesis in terms of the statistical test to be used.
You have already done steps 1 - 4. This section will illustrate step 5: how to formulate and state statistically testable hypotheses.
Hypotheses are predictions. When you hypothesize, you are making an inference. A hypothesis is an assumption about something. It is generally understood as an assumption to be argued, or proved, or tested somehow. It is used in research as an assumption or prediction about a population. Since it is based on a sample of the population, it is tested statistically in order to determine whether the sample really is representative of the population.
A hypothesis makes a statement about the population
In steps 1-4 you determined the variables that will be included in your study, and you determined their relationships in terms of dependent and independent variables. When you set up your research design, you indicated, both graphically and narratively, that relationship.
You set up a theory about values and relationships of those variables as they exist in your population.
Since you intend to survey a sample rather than the entire population, you will make statements (hypotheses) about the population-statements that you will test statistically.
Statistics will tell you whether your statements about the population are valid.
Anyone looking at your research design can see, actually, what you are hypothesizing. However, your research must include the specific hypothesis statements along with your design, and later it must include the tests and test results applied to each statement.
If you look back at your research questions, you will notice that you are interested in values and relationships of the variables of your population:
(1) value: what is the mean of the population research grade?
(2) relationship: are male grades and female grades different? did classification make a difference?
(3) dependence: are "study" and "gender" dependent on one another?
(4) relationship: did grades go down as number of semester hours carried went up?
You will now formulate these values and relationships into statistically testable hypotheses. As preparation to do this, you'll do two things:
You'll start by using the following terminology for (1)-(4) above.
(1) the mean of a population
(2) a difference of means (mean male grade vs mean female grade)
(3) the independence of variables (study and gender)
(4) the correlation of variables (grades and semester hours)
You'll review various types of statistical tests. Learning the terminology and reviewing a few selected statistical tests will help you to formulate and state the hypotheses correctly.
LOTS OF REPETITION—BY DESIGN
STATING THE HYPOTHESES
earlier in this unit you developed a research design showing the variables you selected for your research study. You then expressed the relationships you were predicting by asking research questions. You will now simply express those questions in the form of hypotheses.
A hypothesis is a prediction. Instead of asking (RQ 1)
"What is the mean research grade of all students who have taken research in the last two semesters?" we will state it in the form of a hypothesis-a prediction: the mean grade is --. You will then test the prediction.
In order to test a prediction statistically, you must state it in a precise way and make very clear what variables are involved.
Ok. Let's state your prediction: The mean research grade is 80 or higher (≥ 80).
HERE IS A STATISTICAL BULLETIN:!!!
You don't test your prediction. You test the opposite of your prediction!
The hypothesis that says the opposite of what you are predicting is the null hypothesis (Ho)
The Null Hypothesis
It is more scientific to apply the statistical test to the opposite of your prediction, to try to disprove the opposite. (Some statisticians refer to this as "the good sport theory.") If you can disprove the opposite, then your prediction was right.
This "opposite" hypothesis is referred to as the null hypothesis and is indicated this way:
Ho The mean research grade is less than 8o (< 80)
IMPORTANT!! Test the null Hypothesis
If the null hypothesis is disproved (rejected), then you accept your research (alternate) hypothesis. From now on you will refer to the hypotheses as null and alternate. (You can also refer to the alternate as the research hypothesis). They will be indicated as:
Ho (null) and HA (alt). You can also "number" the hypothesis: H1 or HO1
A hypothesis includes the variable or variables whose relationship is to be tested. You will notice, however, that H1 contains only one variable; you are simply predicting the data on one variable. This will require a univariate (one variable) test. If you refer back to the table at the end of the Operational Definitions, showing "shorthand for using tests," you will see that for a univariate test we will use the ONE SAMPLETTEST, ORTTEST FOR 1 MEAN.
H1 TESTING WORKSHEET
Variable 1: research grade Meas scale: Ratio Type of test: one Sample T Test
The hypothesis for a one variable test can be stated this way:
H0 The mean research grade is less than 80 (<80)
H1 The mean research grade is 80 or greater than 80 (≥ 80)
The rest of your hypotheses (RQ 2 - 8) will require bivariate tests-two variable tests. In each you will be testing relationships. Note that these were the questions asked in your research design.
(Continues...)
Excerpted from CREDIBLE RESEARCH MADE EASYby DORCAS MLADENKA Copyright © 2012 by Dorcas Mladenka. Excerpted by permission of iUniverse, Inc.. All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
