NOTES BASED ON RESEARCH METHODOLOGY BY WELMAN, KRUGER AND MITCHELL, 3RD EDITION. THE AIMS OF RESEARCH ? ? Research is a process that involves obtaining scientific knowledge by means of various objective methods and procedures. The term objective indicates that these methods and procedures do not rely on personal feelings or opinions and that the specific methods are used at each stage of the research process. These methods include procedures for drawing samples (e. g. simple random sampling), measuring variables, collecting information and analysing the information.
Research methodology considers and explains the logic behind research methods and techniques. ? ? Sources of non – scientific knowledge ? ? The sources include authority, opinion of peers, traditions, debating and accidental observation. Under authority, non – scientific knowledge is often merely accepted on the basis of the authority of some or other source. In contrast, with scientific approach, one should check the way in which findings are acquired and not accept them merely because they originate from a so – called expert.
Non –scientific knowledge can also be acquired by asking the opinion of peers. Traditions refer to knowledge which is carried over from one generation to the next. Under debating, people attempt to obtain knowledge and insight by arguing in a seemingly logical manner. Accidental observation – this is where we notice something happening in one situation, but do not investigate the phenomenon in a systematic and planned manner ? ? ? ? Core features of scientific knowledge ? ? ? ? They include systematic observation, control and replication.
Systematic observation – scientific approach requires that we plan an investigation in which we use the results of two groups that have applied the methods. Control – scientific knowledge can be obtained in a controlled manner. Control means that alternative explanations for the obtained results should be eliminated systematically. Replication – it must be possible to replicate the research results. It means that similar results should be obtained by other researchers including other research participants in other circumstances. 1 Quantitative and Qualitative Research Cycles ?
Qualitative implies an emphasis on processes and meanings that are not rigorously examined or measured (if measured at all) in terms of quantity, amount, intensity or frequencies. The aims of qualitative research are to establish the socially constructed nature of reality, to stress the relationship between the researcher and the object of study, as well as to emphasize the value – laden nature of the inquiry. Quantitative research methods do not involve the investigation of processes but emphasize the measurement and analysis of casual relationships between variables within a value – free context. ? Differences between quantitative and qualitative research 1. The purpose of quantitative research is to evaluate objective data consisting of numbers while qualitative research deals with subjective data that are produced by the minds of respondents or interviewees. 2. Quantitative researchers use a process of analysis that is based on complex structured methods to confirm or disapprove hypothesis. Flexibility is limited to prevent any form of bias in presenting the results.
In contrast, qualitative research is based on flexible and explorative methods because it enables the researcher to change the data progressively so that a deeper understanding of what is being investigated can be achieved. 3. Quantitative researchers try to understand the facts of a research investigation from an outsider’s perspective. Qualitative researchers try to achieve an insider’s view by talking to subjects or observing their behaviour in a subjective way. 4. Quantitative researchers try to keep the research process as stable as possible.
They focus on the causal aspects of behaviour and the collection of facts that wouldn’t change easily. Qualitative researchers work with the dynamic and changeable nature of reality. 5. Quantitative researchers control the investigation and structure of the research situation in order to identify and isolate variables. Specific measurement instruments are used to collect data. The approach is particularistic. Qualitative researchers make use of a holistic approach, i. e. , they collect a wide array of data, e. g. , documents, records, photos, observations, interviews and case studies.
In conclusion, the purpose of both quantitative and qualitative research is to try and understand the subject’s point of view. Quantitative researchers do it by means of controlling the situation and using remote empirical and inferential methods. Qualitative researchers, on the other hand, use unstructured interviewing and detailed observation processes to gain better information about the views of the subject. 2 THE RESEARCH TOPIC, PROJECT TITLE AND RESEARCH PROBLEM The research process This involves: ? ? ? ? ? ? Identifying research topic. Defining research problem.
Determining how to conduct the study Collecting the research data. Analysing and interpreting the research data. Writing the report. Identifying research topic ? ? ? ? ? First source of ideas is your own experience and reflection of things around you. Become more aware of current events by keeping up to date with media coverage of worthwhile and interesting phenomena. Become more critical of different viewpoints when they are aired and defended by people in your immediate environment. Read more literature. Listen critically to what people around you are talking about.
What are the current issues in the political, social, economic, environmental and other areas. Guidelines for the selection of a research topic ? ? ? Select a topic that is relevant to your own short-term and medium-term career prospects, whether it be a career in research or any other profession. Select a topic that you find intellectually stimulating and that you are convinced will sustain your interest for a number of years. Select a topic that is researchable in the sense that you will be able not only to merely complete it with the available resources, but also complete it at a level of scholarship that is scientifically acceptable.
Select a topic that you find interesting and worthwhile. Then read more about it and focus more narrowly, delineating it into a more specific topic. ? Defining the research problem ? ? ? ? This involves narrowing down our general interests in a research topic in order to focus on a particular research problem which is small enough to be investigated. This process leads to the setting of the research questions. In academic research, the classic way to identify a research problem is to consider the literature and identify any gaps. These gaps indicate original areas to research. 3 ? ? ? ? ?
A research problem refers to some difficulty that the researcher experiences in the context of either a theoretical or practical situation and to which he/she wants to obtain a solution. To define a problem correctly, the researcher must know what a problem is. To answer a research problem, the researcher must be able to answer the following questions: o What is the problem? o What is the best way to solve the problem? A useful strategy that can be used to identify the research problem involves asking questions. Good questions have the following characteristics: o They express relationship(s) between two or more variables. They are clear, i. e. , what is asked is understood. o Implicit in each research question (research problem) is the view that some variables are the causes of other variables. o The independent variable (X) is that factor which the researcher selects and manipulates in order to determine its effect on the observed phenomenon. o The variable is independent because the researcher is interested in how it affects the other variables being studied. o The dependent variable (Y) is that factor which the researcher observes and measures to determine how it was affected by the independent variable.
The origin of research problems Typical sources of research problems include: ? ? ? Practical problems – caused by political, social and economical changes. Previous research – from other studies where there are shortcomings or contradictions. Theories – should associate a research project with a specific theory. A theory is a statement or a collection of statements that specify the relationships between variables with a view to explaining phenomena. The purpose of research ? ? ? To describe how things are, i. e. , define the nature of the study object. To explain how things are the way they are.
Could it be because one thing caused another? To predict phenomena. Research hypothesis ? A hypothesis is a statement or proposition that can be tested by reference to the empirical study. Example, a research question may be: What is the relationship between advertising expenditure and income? 4 A hypothesis dealing with the same question is: There is a positive relationship between advertising expenditure and income. ? ? ? A research question differs from a hypothesis in that the research question is always expressed as a question, while a hypothesis is expressed as a statement.
Hypothesis can either be directional or non-directional. Non-directional hypothesis do not predict the direction of the difference or relationship. o There is a difference between the organisational commitment of male and female employees. o There is a relationship between job satisfaction and salary level. Directional hypothesis are used if the researcher is more confident about the direction of the difference or relationship, or if the literature reports that previous studies found differences to be in a particular direction. Female employees have a higher level of organisational commitment than male employees. o There is a positive relationship between job satisfaction and salary level. A null hypothesis states that there is no difference between two groups in relation to some variable, or that there is no relationship between two variables. o There is no difference between the organisational commitment of male and female employees. o There is no relationship between job satisfaction and salary level. A null hypothesis is indicated by H0 whilst the alternative hypothesis is indicated by H1. E. g. H0: There is no relationship between job satisfaction and salary level. H1: There is a positive relationship between job satisfaction and salary level. ? ? ? Formulating the research hypothesis ? Where possible, we should translate the research problem into a research hypothesis that states a relationship between two or more variables in one (or more) population(s). ? Use the following approach to develop a hypothesis: o Discuss the problem, its origin and the objectives in seeking a solution with experts. o Examine data and records concerning the problem for possible trends and clues.
Use secondary data. o Interview interested parties and individuals on a limited scale to gain greater insight into the practical aspects of the problem. 5 NOTES ON LITERATURE REVIEW ? ? Literature review is a review of the existing scholarship of available body of knowledge to see how other scholars have investigated the research problem you are interested in. You learn from other authors how they theorised and conceptualised on issues, what they have found empirically, what instrumentation they have used and to what effect. Elements of Literature review ? ? ? Definitions Different theories, models and hypothesis in the field of research Existing data and empirical findings that have been produced by previous research. Measuring instruments (questionnaires, scales and indices) that have been developed to measure the extent or scope of research. Importance of Literature Review ? ? ? ? ? ? ? To ensure one does not merely duplicate a previous study. To discover what the most recent and authoritative theorising about the subject is. To find out what the most widely accepted empirical findings in the field of study are. Researcher can relate those findings to his/her own study.
To identify the available instrumentation that has proven validity and reliability. To ascertain what the most widely accepted definitions of key concepts in the field are. Provides important facts and background information about the subject under study. Researcher can get ideas on how to proceed with the investigation. Planning the Literature Search ? ? ? Plan carefully to locate relevant and up-to-date literature. Define the parameters of your search – guided research questions and/or objectives. Generate key words and search terms – this is the most important part of planning your search for literature.
Key words are the basic terms that describe your research questions and/or objectives. Discuss your ideas as widely as possible – with experts and peers. ? Compiling a Literature Review ? Integrating the studies o Literature review should not consist of mere compilation of separate, isolated summaries of individual studies of previous researchers. o You should clearly show how these studies relate to one another and how the proposed research ties in with them. Sources of Literature Searches Primary Literature Sources 6 ? ? ? o These are the first occurrence of a piece of work.
They include published sources such as reports and some central and local government publications such as economic surveys, UNESCO reports, the Hansard etc. o They also include unpublished manuscripts such as letters, memos and committee minutes that may be analysed as data in their own right. Secondary Sources They include books and journals that constitute the subsequent publications of primary literature. Tertiary Sources o These are designed either to help locate primary and secondary literature or to introduce a topic. o They include indexes and abstracts as well as encyclopaedias and bibliographies.
Search using CD-ROM A CD-ROM contains permanent digitally encoded information on a large on a large scale which may represent text, graphics, images or data and can be accessed very quickly. Examples of CD-ROM databases include: o ABI/INFORM – Covers every aspect of business management theory and practice including accounting, computer, human resources, marketing and organisational behaviour. o Anbar – Includes subjects such as accounting and finance, information management and technology, management services and production, marketing and distribution, personnel and training, and top management. ? ? ? ? ? ? ?
Use of Internet for example Google Scholar, Science Direct, Elsevier etc. Quotations o With exception of direct quotations, the entire research report should be presented in the author’s own words i. e. without paraphrasing or patching together pieces from other sources. o Use direct quotations sparingly as they are only permitted when something is expressed so eloquently or in such an original way that you feel something will be lost in the process of reformulating it. General guidelines for using quotes in research reports o Incorporate quotations of up to (and not longer than) about 30 words into the text. Enclose the quoted material in double quotation marks. o Apart from the surname of the author and the year of publication, all quotations should also be accompanied by the page number on which the material appears in the quoted source. 7 ? Evaluating the relevance of the literature o How recent is the item? o Is the item likely to be superseded? o Is the context sufficiently different to make it marginal to your research question(s) and/or objectives? o Have you seen references to this item (or its author) in other items that were useful? Does the item support or contradict your arguments? o Does the item appear to be biased? o What are the methodological omissions within the work? o Is the precision sufficient? ? ? ? ? ? ? The reference system A reference must enable the reader of your research report to locate the information sources referred to if so needed. A reader may want to read more about the issue from the information source. References in the text If you refer to theories, research findings, or any other contribution previously reported, you should give the sources involved due credit.
Give the surname(s) of the author(s) followed immediately by the year of publication between parentheses. E. g. In a well-designed experiment, Strauss (1990) found that… Do not list a string of references that are not appropriate in the context in which you cite them. Cite sources after identifying their relevance. It should be clear whether the citation is a finding or an opinion that has been obtained from the source. If a source has more than two authors, you should list them all the first time it appears in the report. E. g. Ojah, Muhanji and Myburg (2008) found that…
In all subsequent references only the surname of the first author is provided followed by et al. E. g. Ojah et al (2008) found that… If you refer to the same source more than once in the same paragraph, the date should accompany only the first reference. If there is more than one source with several authors, and they have the first two authors and date in common, give the first three names plus et al in further references to distinguish them. E. g. Smith, Jones, Botha and Tiffin (1990)… 8 ? ? ? ? ? ? ? ? ? ? ? Smith, Jones, De beer and Zwane (1990)…
Further references are Smith, Jones and Botha et al (1990)… and Smith, Jones and De Beer et al (1990)… When more than one reference appears between brackets, put a comma between the author(s)’ surname(s) and date, and a semicolon (;) between the different references. E. g. (Jaffe, 1974; Seyhun, 1986; Pope et al, 1990) To distinguish between publications of the same author published in the same year, alphabetise them according to their titles and affix the letters a, b, c and so on onto the dates. E. g. IMF and IDA (2006a), “Heavily Indebted Poor Countries (HIPC) Initiative – Statistical Update.
March 21, 2006. IMF and IDA (2006b), “Heavily Indebted Poor Countries (HIPC) Initiative – List of RingFenced Countries that meet the Income and Indebtedness Criteria at end 2004. IMF and IDA (2006c), “Heavily Indebted Poor Countries (HIPC) Initiative – Issues Related to the Sunset Clause. August 16, 2006. Distinguish between different authors with the same Surname by Providing their Respective Initials Together with their Common Surname. E. g. E. O. Abuya (2010) … N. A. Abuya (2010). References to Institutions are done with Minimum Identification or Description.
E. g. (World Bank Economic Report, 2009, p. 20) When a reference is made to a quotation found in a secondary source, it is cited in the text as follows: E. g. (McCallum and Nelson, as quoted by Muhanji, 2010, p. 560) When the date of a publication is unknown, it is cited as follows: According to Ryan and Bernard (n. d. )… Sources in the Reference List Arrange references alphabetically in terms of the surname of the first author. Arrange sources with the same first author but with different co-authors alphabetically according to the surname of the second author.
If the same author or collection of authors has different publications, list them chronologically according to the date of publication. When referencing journal articles, the name of the journal appears in italics (or is underlined). When referencing a book, the name of the book appears in italics (or is underlined). For references from an internet source, indicate the URL and the date when the document was accessed. E. g. IMF & World Bank. (2006). Applying debt sustainability framework for lowincome countries post debt relief. http://www. imf. org/external/np/pp/eng/2006/110606. pdf. (20/11/2007) 9 ? ? ? ? ? ? ?
NOTES ON POPULATION AND SAMPLING ? ? ? ? When we conduct a research, we collect data from our objects of inquiry in order to solve the problem concerned. A crucial element in data collection is the research design that we use. A research design is a plan according to which we obtain research participants (subjects) and collect information from them. In a research design we specify: o The number of groups that should be used (this is necessary to decide which statistical technique to use). o Whether the groups are to be drawn randomly from the population involved and whether they should be assigned randomly to groups. What exactly should be done with them in the case of experimental research. Population and Sample ? ? ? ? ? ? The population is the study object and consists of individuals, groups, organisations, human products and events, or the conditions to which they are exposed. Population encompasses the total collection of all units of analysis about which the researcher wishes to make specific conclusions e. g. all universities, all banks and so forth. A population is the full set of cases from which a sample is taken. The size of the population is indicated by N. A sample represents a part of the population and is denoted by n.
To calculate an appropriate sample, you can use the formula: Where N is the population and Sampling is the level of precision. 1. It is divided into probability and non-probability samples 2. Examples of probability samples includesimple random samples, stratified random samples, systematic samples and cluster samples. 3. Non-probability samples include accidental/incidental samples, quota samples, purposive samples, snowball samples, self-selection samples and convenience samples. 4. In the case of probability sampling, we can determine the probability that any element or member of the population will be included in the sample.
In non-probability sampling by contrast, we cannot specify this probability. 5. Advantages of probability sampling is that it enables us to indicate the probability with which sample results (e. g. sample means) deviate from the corresponding population values (e. g. population means). 6. Unlike non-probability sampling, probability sampling enables us to estimate sampling error. Nevertheless, non-probability sampling is frequently used for reasons of convenience and economy. 10 The Sampling Frame ? This is a list of units comprising a population from which a probability sample is selected. E. g. if a sample of students is selected from a student’s roster, the roster is the sampling frame. Likewise, if a sample is selected from a telephone directory, the telephone directory is the sampling frame. In order to compile an appropriate sampling frame, bear in mind the following checklist: o Are the cases listed in the sampling frame relevant to your research topic? o Does the sampling frame include all cases i. e. is it complete? o Does the sampling frame exclude irrelevant cases, in other words, is it precise? o Can you establish and control exactly how the sample will be selected (for purchased lists). Sampling Methods Probability Sampling 1. Simple Random Sampling a. Each member of the population has an equal chance of being included in the sample and each sample of a particular size has the same probability of being chosen. b. Steps to follow when drawing a random sample include: i. Identify all the units of analysis in the sampling frameand give them cosecutive numbers e. g. 001, 002, 003, and so forth. ii. The mechanism used to choose the unit of analysis should ensure that each member has an equal chance of being selected. This can be done by means of a table of random numbers. 2.
Stratified Random Sampling a. Strata are clearly recognisable, non-overlapping subpopulation that differs from one another in terms of a specific variable. Each subpopulation is called a stratum. E. g. male and female, private and public universities etc. b. In order to draw a representative stratified random sample, the following two aspects must be put in mind: i. Identify the various strata according to one or more variables. ii. Draw a random sample from each separate stratum. c. Advantages of stratified random sample include: i. In a random sample from a normal population that is stratified, e. . in terms of gender, the probability of a sample consisting of only one gender is zero. ii. Stratified random sampling requires smaller samples than simple random sampling in order to obtain valid results. 11 3. Systematic Sampling a. Suppose we need to obtain a sample of n members from a population of N elements (units of analysis) that are numbered from 1 to N. In systematic sampling, we include every element (where, is an integer i. e. a whole number). E. g. if one wants to select 10 schools from a population of 100, draw the first element at random and then select every 10th element.
If the first element is 8, subsequently select 18, 28, 38, and so on. b. Systematic sampling is more practical baecause it requires less time and is cheaper than random sampling. 4. Cluster Sampling a. In large-scale surveys, it is usually difficult, if not impossible to obtain lists of all the members (units of analysis) of the population. b. In the case of cluster sampling, we first draw (stratify randomly) pre-existing heterogeneous groups called clusters c. Randomly select units of analysis from each cluster. E. g. suppose we wish to conduct a survey on primary schools in Kenya. First draw a sample from the 8 provinces.
Then draw a number of districts from each province. Next you draw a number of locations from the district. Finally select a few schools within the location. d. Its disadvantage is that it requires a great deal of time and money to reach one or two individual units in a remote area. If the area is not remote, cluster sampling saves time and money compared to simple random sampling. Summary Sampling Technique Sampling Frame Required Size of Sample Needed Better with a sample size of a few hundred. Suitable for all sizes. Geographical Area Suited Relative Cost Easy to Explain to Support Workers Relatively difficult to explain.
Advantages Compared to Simple Random Sampling Simple Random Accurate and easily accessible. Systematic Stratified Random Accurate, easily accessible. And not containing periodic patterns. Actual list not always needed. Accurate, easily accessible, divisible into relevant strata. Concentrated if face-to-face contact required, otherwise does not matter. Concentrated if face-to-face contact required, otherwise does not matter. High if sample size is large or if sampling frame is not computerised. Low Relatively easy to explain. Normally difference. no Suitable for all sizes.
Concentrated if face-to-face contact required, otherwise does not matter. Low, provided that lists of relevant strata are available. Relatively difficult to explain. Once strata is decided, it is easy to explain. Better comparison across strata. Differential response rates may necessitate re-weighing. 12 Cluster Accurate, easily accessible, relates to relevant clusters not individual population members. As large as practicable . Dispersed if faceto-face contact is required and geographicallybased clusters are used. Low, provided that lists of relevant clusters are selected.
Relatively difficult to explain until clusters are selected. Quick precision reduced. but is Non-Probability Sampling ? ? Advantage of non-probability sample is that they are less complicated and more economical (in times of time and financial expenses) than probability samples. Non-probability samples may be especially useful in pilot studies in which a preliminary form of a questionaire has to be tested. 1. Accidental/incidental Sampling o An accidental sample is the most convenient collection of members of the population (units of analysis) that are real or readily available for research purposes. E. g. people who happen to be at a certain shopping centre at a particular time or organisations that are close to the researcher’s home. 2. Quota Sampling o In this method, the researcher makes an effort to have the same proportions of units of analysis in important strata such as gender, age, and so on as are the population. o Quota sampling yields quite satisfactory results under the following circumstances: ? One should know the important strata according to which the population is composed and their respective proportions with a reasonable degree of accuracy, or be able to estimate them. One should include enough cases (at least 15, but preferably more than 25) of each stratum in the sample. 3. Purposive Sampling o Researchers rely on their own experience, ingenuity and/or previous research findings to deliberately obtain units of analysis in such a manner that the sample they obtain may be regarded as being representative of the relevant population. o The problem with this kindof sampling is that different researchers may proceed in different ways to obtain a sample. o It is impossible to evaluate the extent to which samples are representative of the relevant population. . Snowball Sampling o First approach a few individuals from the relevant population 13 o These individuals then act as informants and identify other members (e. g. acquaintances or friends) from the same population for inclusion in the sample. o The latter may, in turn, identify a further set of relevant individuals so that the sample, like a rolling snowball, grows in size till saturated. 5. Self-Selection Sampling o This occurs when we allow a case, usually an individual, to identify their desire to take part in the research. o The researcher will do the following: ?
Publicise the need for cases, either by advertising through appropriate media or by asking them to take part. ? Collect data from those who respond. o Cases that self-select often do so due to their feelings or opinions about the research question(s) or stated objective(s). 6. Convenience Sampling o Convenience or haphazard sampling involves selecting haphazardly those cases that are easiest to obtain for the sample, such as, the person interviewed at random in a shopping centre or airport or university. o The sample selection process is continued until we reach the required sample size. This technique is borne to bias and influences that are beyond control due to the fact that the cases appear in the sample because they were easy to obtain. o Subsequent generalisations are likely to be flawed at best. o Such samples serve best as pilot studies using more structural samples. Sample Style Likelihood of Sample being Representative Very low Type of Research in Which Useful When performing explorative research. When you do not expect to generalise findings. Where costs are constrained or data needed very quickly so that an alternative to probability sampling is needed.
Where working with very small samples. Focus: unusual or special, key themes, in-depth, importance or case, illustrative, Relative Costs Low Control Sample Contents Low over Accidental (Incidental) Quota Purposive Snowball Self-Selection Convenience Reasonable to high, although dependent on selection of quota variable Low, although dependent on research choices: extreme case, heterogeneous case, homogeneous case, critical case, typical case Low, but cases will have characteristics desired. Low, but cases self selected. Very low Moderately high to reasonable Reasonable
Relatively high Reasonable Where difficulties in identifying cases. Where explorative research is needed Where very little variation in population reasonable Low Low Quite low Low Low 14 TYPES OF RESEARCH DESIGN 1. 2. 3. 4. Experimental research Quasi – experimental research Non – experimental research Qualitative research The first three constitute quantitative research. 1. Experimental Research ? This involves some form of intervention, i. e. , the participants (units of analysis) are exposed to something to which they would not have been subjected to. In the hypothesis we express the influence that the independent variable is expected to have on the dependent variable and it is this influence we measure in the experiment. ? We therefore measure the extent to which the intervention (independent variable) has changed or affected the unit of analysis (dependent variable). Measuring the dependent variable Unit of analysis Intervention ? ? ? ? ? The measurement of the dependent variable before the intervention is called premeasurement and the measurement after the intervention is called postmeasurement.
When we work with one group only, the possibility exists that considerable changes in the dependent variable could have occured without the intervention (independent variable). We therefore need a control group that is not exposed to the experimental intervention to which we can compare the group that was exposed to the intervention. A control group therefore is a group that does not receive the intervention, but serves to exercise control over the nuisance variables. A nuisance variable is any variable which is not mentioned in the research hypothesis that may influence the dependent variable.
Characteristics of true experimental research a. Control over the independent variable – This means that we may determine which levels of the variable we should use. b. Random assignment of units of analysis to groups i. Different groups are formed by means of random assignment. ii. The term random refers to the way in which the groups are assigned to the different levels of the independent variable. c. Nuisance variables (or third variables) The basic purpose of experimental research is to control nuisance variables to such an extent that the various groups differ only in terms of the levels of the independent variable in question. 5 Ways of controlling the influence of nuisance variables i. Determine the speculative effect of the nuisance variable by doing a proper literature review. In this way, nuisance variables documented in earlier research findings can be identified and controlled. ii. In few cases, it may be possible to eliminate the effect of the nuisance variable completely. iii. Build nuisance variable into the design as an additional independent variable. iv. Purposely form the various groups so that they are as similar as possible in terms of all variables except the independent variable.
Causality in the human behavioural sciences. A variable X may be regarded as a cause of another variable Y if each of the following three conditions are met. ? ? ? There must be a correlation between the variables. The cause must precede the effect. The third variable must be controlled. Correlation between variables ? ? ? A correlational association between X and Y exists when Y tends to appear in association with X. E. g. correlation between salary earned and productivity. The existence of a correlational effect does not mean that X causes Y.
A correlation between X and Y may be the result of either a mutual relationship or a third variable Z, which may cause both X and Y. E. g. a high positive correlation between number of churches and the incidence of crime. This correlation may be attributed to a third variable, size of the population. A mutual relationship means that one variable (X) plays a role in the occurence of another variable Y which, in turn, affects the former so that there is a mutual relationship between them. From example above, the size of the population has an effect on both the number of churches and the incidence of crime. Cause must precede effect A variable or event X is considered to be a necessary cause of Y if Y cannot occur in the absence of X. E. g. motivation increases productivity. 2. Quasi-experimental Research a. True experimental research is impossible in business and administrative sciences. b. Quasi – experimental research differs from the experimental research in that the researcher cannot randomly assign subjects to the different groups. 16 There are two groups of quasi-experimental research: a. The non-equivalent control group design.
Here we use two pre-existing groups as an experiment and control group, respectively. i. The threat of group differences that may already have been present prior to the start of the experimental intervention is taken into account by measuring both groups on the dependent variable prior to the experimental observation. ii. If the experimental and control groups do not differ in terms of the premeasure but do differ in terms of the postmeasure, we can ascribe the the difference in the postmeasure to the difference in intervention that they hsve received.
Premeasure Experimental group Control group Intervention Postmeasure Experimental group Control group b. The interrupted time-series design i. In this, more than one measurement of the dependent variable is obtained, with equal intervals both before and after the intervention. ii. Researcher may plan the intervention or it may be an unplanned event like collapse of Lehmann brothers. 3. Non-experimental Research a. Neither random assignment nor any planned intervention occurs in nonexperimental research. b.
In this type of research one or more variables, apart from the independent variable in question, could be the actual cause of observed variation in the dependent variable. E. g. , knowledge about astronomy has been obtained by non-experimental means because we cannot manipulate the stars. c. Because there is no planned intervention and the data are collected in a field situation, this design is supposed to show the greates similarity to real life. d. Research designs include: survey, non-experimental research designs involving measurements at a single time, longitudinal and opinion polls. 7 VALIDITY OF CONCLUSIONS ? ? We test research hypothesis to help us decide whether or not a specific implication inferred from a theory is tenable, or to provide an answer to a research question. The conclusion we reach about the relationship should be a true reflection of the relationship. E. g. if we wish to investigate whether attitudes have changed over time, a design that selects data at only one point in time would be inappropriate. Validity therefore refers to the extend to which an empirical measure adequately reflects the real meaning of the concept under onsideration. Internal Validity ? ? ? Internal validity describes the degree to which changes in the dependent variable are indeed due to the independent variable rather than to something else. To promote internal validity, it is necessary to eliminate all possible threats to internal validity. If we allow the threats to operate unchecked, we cannot unequivocally interpret observed changes in the dependent variable as the effect of the independent variable, and this will render the internal validity of our conclusions subject. Sources of internal invalidity/ threats to internal validity 1.
History – During the cause of the experimental research, historical events may occur that will confound the results. 2. Maturation – People are continually growing and changing, and such changes can affect the results of the experiment. In long-term experiments, people grow older and wiser. In short-term experiments, they may grow tired, hungry, sleepy, bored and/or change in other ways that affect their behaviour in the experiment. 3. Testing – The process of testing and retesting influences people’s behaviour, thereby confounding the experimental results.
E. g. , suppose we administer a questionnaire to a group as a way of measuring their prejudice. Then we administer an experimental stimulus and re-measure their prejudice. By the time we conduct the post-test; the subjects may have become more sensitive to the issue of prejudice and will be more thoughtful in their answers. 4. Instrumentation – The process of measurement in pretesting and post-testing brings in some issues of conceptualisation and operationalisation (operationalisation means putting a variable in measurable terms). E. g. if we used different measures of the dependent variable in the pre-test and post-test (e. g. different questionnaires about prejudice), how can we be sure they are comparable to each other. External validity ? ? ? This refers to the possibility that conclusions drawn from experimental results may be generalizable to the real world. Population validity refers to the degree to which the findings obtained from a sample may be generalised to the total population to which the research hypothesis applies. Biased results may be obtained if data are obtained from unrepresentative samples. 8 DATA-COLLECTION METHODS AND MEASURING INSTRUMENTS IN QUANTITATIVE RESEARCH a. Systematic observation and quantitative measurement Systematic observation means that independent observers other than ourselves should also be able to observe and report whatever we, as researchers, observe and report. b. Measurement theory ? ? ? ? If we wish to investigate the accuracy or “truth” of the research hypothesis, we need to measure the dependent variables appearing in it. Measuring variables such as length or mass of things or people (subjects) does not present many problems.
In human behavioural sciences, the variables we wish to measure are constructs such as liquidation, attitude towards affirmative action, entrepreneurship, helping behaviour etc. We measure these constructs indirectly by means of indicators. E. g. , a variable like socioeconomic status is a construct because there is no single indicator that perfectly encapsulates it. To measure socio-economic status we may use income, occupation, educational level, area in which one resides and so on. The nature measurement ?
Measurement involves the assignment of numbers, in terms of fixed rules, to individuals (or objects) to reflect differences between them in some or other characteristic or attribute. The rules in terms of which numbers are assigned constitute an operational definition of the variable being measure e. g. length. In measurement we distinguish between different levels of measurement on the basis of the following characteristics: o Distinguishability (the number 3 is different from the number 2). o Order of rank (2 has a higher rank than 1) o Equal intervals between successively higher numbers .
Levels of measurement include nominal, ordinal, interval and ratio. i. Nominal measurement ? ? ? ? In this case, the numbers we assign to individuals only serve to distinguish them in terms of the attribute being measured. In this kind of measurement, we place individuals in different, mutually exclusive and exhaustive categories in respect of such characteristics. By mutually exclusive we mean that each person belongs to one of the categories only e. g. male or female. Exhaustive implies that all individuals can be accommodated in some or other category. E. g. we could measure someone’s occupation by assigning the number ? ? 19 1 to doctors, 2 to lawyers, 3 to teachers, 4 to accountants etc. The number 4 does not in any way imply a greater magnitude of the variable occupation. ii. Ordinal measurement ? ? In this case, the numbers we assign not only reflect differences among individuals but also rank order. The assumption is made that those to whom we assign higher numbers exhibit more of the particular attribute than those to whom we assign lower numbers. E. g. suppose the order of rank of administrative staff in a company is as follows: assistant administrative officer, administrative officer, senior administrative officer.
We therefore assign the numbers 1 to 3 as follows: 1 to the rank of assistant administrative officer, 2 to the rank of administrative officer and 3 to the rank of senior administrative officer. iii. Interval measurement ? An interval scale has all the characteristics of both nominal and ordinal scales, but provides additional information regarding the degree of differences between individual data items within a set or group. ? In this measurement, therefore, we use the property of equal differences between consecutively higher numbers. E. g. trust in a company’s future; the question is: This company will make a profit till 2020. Absolutely 1 2 3 4 5 Not at all. iv. Ratio measurement ? ? ? ? Ratio scales represent the highest level of precision. In this, there is fixed and absolute zero point. The ratios between numbers assigned to this level of measurement can therefore be interpreted meaningfully. The advantage of using ratio scales is that any statistical analysis can be used on data collected in this form. e. g. , Net income from sales for a small scale business. 10000 20000 30000 40000 c. Validity ?
Recall that validity is the extent to which the research findings accurately represent what is really happening in the situation. ? Any given measuring instrument measures the construct intended and random measurement error (reliability). 20 ? Types of validity that relate to the independent variable include: construct validity and criterion-related validity. Construct validity ? Construct validity of the operationalisation of an independent variable is the degree to which the procedures intended to produce the independent variable of interest succeed in generating this variable rather than something else. In practice, we seldom investigate the construct validity. ? Usually we clearly describe the way in which such a variable has been operationalised and leave it to the reader to judge whether these operations have succeeded in bringing about the desired independent variable. ? The construct validity of a measuring instrument refers to the degree to which it measures the intended construct rather than irrelevant constructs or measurement error. ? It is advisable to use more than one measure of the same construct.
If this is not done, it is impossible to examine to what extend any given measuring instrument measures anything else but itself. ? An important threat to the construct validity of questionnaire measurements of personality, interests, and attitudes lies in measurement reactivity. ? The response sets of faking and social desirability and the response style of acquiescence imply that individuals’ responses are not true reflections of their personalities, interests, and attitudes. ? Faking means that participants deliberately distort their responses in order to create a desired impression. E. g. question like do you easily lose your temper? A respondent will answer negatively because he wishes o create a good impression. ? In social desirability, subjects deliberately or inadvertently provide the responses that they believe to be socially accepted. E. g. a question like do you like everybody you know? Respondents may answer positively because they believe that it would be socially unacceptable not to like some people. ? Acquiescence refers to the phenomenon where research participants tend to consistently answer “yes” (to no/yes items) or “true” (rather than false), irrespective of the content of the question.
E. g. an interviewee may answer “yes” to both of the following statements: I prefer to spend a night out in town with friends instead of staying at home reading a book. I prefer a quiet night at home reading a book instead of partying with friends. Answering yes to both questions is contradictory. Criterion-Related Validity ? This refers to the degree to which diagnostic and selection measurement/tests correctly predict the relevant criterion. 21 ? ? The relevant criterion refers to the variable that is to be diagnosed or on which success is to be predicted respectively. E. g. he validity of university exams is shown in their ability to predict students’ success in university. d. Reliability ? ? ? This is concerned with the findings of the research and relates to the credibility of the findings. In determining whether our findings are reliable, we need to ask the following question: will the evidence and conclusions stand up to the closest scrutiny? It stands to reason that if we measure a construct (e. g. small business success) by means of a particular instrument (e. g. taxable profit), comparable measurements should be obtained for the same individual/objects irrespective of, e. . , when the instrument is administered, which particular version of it is used, and who is applying or administering it. Reliability therefore refers to the extent to which the obtained scores may be generalised to different measuring occasions, measurement/test forms and measurement/test administrators. Generalisation refers to the consistency of the ranking (of scores) that we assign to the individuals or objects, irrespective of when the measuring instrument was applied, which form it was used, and by whom it was administered.
Scores that are assigned to individuals should therefore be consistent irrespective of the time of measurement, the test used, and the person administering the test. ? ? ? Estimating reliability ? If a research finding can be repeated, it is reliable. In other words, if anyone else were to repeat the research, they should be able to obtain the same results as those obtained originally. E. g. , if we find that a group of workers who attended a training course doubled their previous productivity levels, and another researcher obtained similar results, the measurements are reliable.
Repeating a research study in order to establish reliability is known as replication. Test-retest reliability ? To determine test-retest reliability of a measuring instrument, we should administer it on at least two occasions to the same large, representative sample from the population for which the instrument is intended. We then correlate the two sets of scores obtained in this way and calculate a correlation coefficient for the two sets of data. This coefficient is an index of reliability. 22 ? ? ? ? ? ? ? ?
The time interval between the two administrations should not be too long to prevent real and permanent changes from taking place in the attribute being measured. At the same time, it should not be too short so that participants may remember the responses given on the first administration. A disadvantage of this method is that it is often difficult to persuade respondents to answer questions a second time. Furthermore, if they do answer the questions a second time, they may think more deeply about the questions on the second occasion and provide different answers. Parallel-forms reliability ?
Parallel-form reliability of a measurement test is determined by using interchangeable versions of a measurement/test that have been compiled to measure the same construct equally well but by means of different content. The different versions are administered to the same representative sample and the obtained scores are correlated. Parallel-forms reliability deals with the generalisability over parallel measurement/test forms. ? ? Internal consistency ? ? A high internal consistency implies a high degree of generalisability across the items within the measurement test.
In other words, if a person performs well on a few items in such a measurement/test, the chances are good that he/she will fare equally well on the remaining items in the measurement/test. Every item is correlated with every other item across the entire sample and the average inter-item correlation is taken as the index of reliability. To determine the reliability of a measuring instrument by using the internal consistency method, we administer the measurement/test only once to a large representative sample.
It is important to note that the longer a measurement/test is, in other words, the more appropriate content it covers, the higher its internal consistency should be. ? ? ? Split-halves reliability ? ? This is calculated by correlating the scores on one half of the test with the scores for the other half of the test. The test is therefore divided into equal halves which are the correlated in terms of the scores for each half. 23 ? The correlation coefficient for the two halves is then calculated and indicates the reliability of the test.
In conclusion, a good measurement technique should be both valid (measuring what it is intended to measure) and reliable (yielding a given measurement dependability). Reliable but not valid Valid and not reliable Valid and reliable Pilot Studies in the development of a questionnaire ? ? When a new measuring instrument is developed, it is useful to “test it out” before administering it to the actual sample. The process of “testing out” is done by means of a pilot study which entails administering the instrument to a limited number of subjects from the same population as that for which the eventual project is intended.
The pilot study is therefore a “dress rehearsal” for which the eventual project is intended. ? Purpose of a pilot study ? To detect possible flaws in the measurement procedures such as ambiguous instructions, adequate time limits as well as in the operationalisation of the independent variables. E. g. , you may have to use a range of income instead of a fixed numerical value. To identify unclear or ambiguously formulated items. In such a pilot study, the actual questions are put to the respondents, and they are then asked to indicate how they have interpreted the formulated questions.
An opportunity for researchers and assistants to notice non-verbal behaviour on the part of the participants that may possibly signify discomfort or embarrassment about the content or wording of the questions. NB: it is virtually mandatory to test survey questionnaires on a small group of individuals who are representative of the populations for which they are intended. If the instrument is extensively revised in reaction to the results of the pilot study, the revised instrument should be subjected to a new round of testing. ? ? 24 Measuring instruments ? ? ? Sources of data include primary data and secondary data.
Primary data are original data collected by the researcher for the purposes of his/her own study at hand. Secondary data are information collected by individual or agencies and institutions other than the researcher him – or herself. Types of measuring instruments 1. Survey questionnaires, standardised measuring instruments and attitude scales. These do not involve direct observations of the behaviour of subjects but, instead, individuals’ reports of their behaviour. These measuring instruments are therefore susceptible to measurement reactivity (like withholding of co-operation to deliberate deception). 2.
Rating scales which require raters to assess the behaviour of participants. Bias on the part of the raters can invalidate such information. 3. Indicators such as the inflation rate, CPI, retail sales, registered unemployed, FDI, etc. These are compiled and calculated by people such as government officers. The information (figures) is published in reports, manuals, databases, etc. Unobtrusive measurement ? ? ? This occurs when a subject’s behaviour is observed and the participant is unaware that measurement is taking place. The researcher can observe through either one-way mirror, hidden cameras or behave like a detective.
The problem of measurement reactivity is completely eliminated. Group contacts Advantages of collecting data from groups of people 1. Captive audiences such as students or prisoners, the elderly in a home for the aged etc are available to work with. This procedure then corresponds to the administration of a group measurement/test. 2. Since a single person (with possibly a few assistants) is required to provide the instructions in one room or hall, the cost per questionnaire (in the case of surveys) is much lower than that of a personal interview. 3. The researcher is in full control of the completion of the questionnaires.
The session is arranged with the permission of the appropriate authorities (e. g. schools, university etc) so that no respondent has an excuse for not completing the questionnaire. Response rate is close to 100%. 4. Since the researcher and his/her assistants are present, queries about the completion of survey questionnaires may be answered immediately. 25 5. The group contact way of administering questionnaires corresponds to the personal interview as far as the presence of the interviewer is concerned, but it allows for the same degree of anonymity as the typical postal survey. ? ?
The disadvantage of this method is that it is limited to a few populations. Categories of group contacts include: survey questionnaires and postal dispatch, standardised tests, attitude scales, and rating scales and situational tests. Survey questionnaire and postal dispatch ? Survey questionnaires are used to obtain the following types of information from respondents: ? Biographical details e. g. age, educational qualifications, income etc. ? Typical behaviour e. g. the television programme they favour. ? Opinion, beliefs and convictions about any topic or issue e. g. the present political state. ? Attitudes e. g. towards affirmative action.
An example of a survey questionnaire is: The leadership in my organisation at the present time is: Hardworking ? ? 1 2 3 4 5 Lazy ? ? ? ? ? Attitudes should preferably be assessed by means of attitude scales rather than survey questionnaires. Whereas attitude scales are completed by the respondents typically under the supervision of research staff, we can obtain information about biographical particulars, typical behaviour, opinions, and beliefs in persons, by telephone, mail or email. When compiling the questionnaire, we should take the eyesight and the literacy level of the intended respondents into consideration.
Not only should the intended respondent be able to read and write, but they should also be able to follow instructions. Ignorant respondents may find complicated filter questions, which involves different routes for responses confusing. The purpose of filter questions is to determine whether her respondents should answer all subsequent questions or whether they can omit some of them. Unlike personal interview and telephonic interviews, the respondents cannot fall back on anybody else but themselves since the interviewer is not available to direct them around irrelevant questions.
Conducting a typical postal or mail survey involves the following two stages: ? We first assemble the questions asking for information about biographical particulars, typical behaviour or opinion and beliefs that we want to put to the respondents in a structured questionnaire. 26 ? We then post the questionnaire to respondents with the request that they be completed and returned by mail or fax. Advantages of postal surveys 1. Cost and ease of application A postal survey is the least expensive of all survey methods.
Irrespective of how far and wide the respondents are scattered across the country, they can all be reached by means of a relatively low postal cost. 2. Anonymity Postal survey provides the greatest possibility of anonymity, i. e. , no name or identification is given. Disadvantages 1. Control over responding The researcher has least control over the conditions under which postal questionnaires are completed. The chances are great that some questions may be omitted or not be respondent to in the order presented, or someone else may complete the questions. 2.
Response rate Researchers lack control over the completion of the questionnaires. It may result not only in poorly completed questionnaires, but also a poor response rate. Postal survey response rates frequently fall below 50% of the target population. To improve the response rate of postal surveys, the questionnaire may be: ? ? ? ? Delivered in person to the respondents’ addresses with a request that they be posted back on completion. Posted and collected personally. Both delivered and collected personally. Questionnaires can be followed up by letters and postcards reminding the respondents to complete and return them.
Standardised tests ? A standardised test is a collection of tasks in which the content, the administration and the scoring of the obtained responses are the same, irrespective of who is administering it and by whom it is scored. Standardised tests take the form of individual test and group tests. Individual test may only be administered individually, in other words, a test administrator can administer it to only one testee at a time. Group test may be administered to more than one individual simultaneously in one session.
Types of tests include aptitude tests (which include intelligence tests), academic achievement tests, personality tests, and interest tests. 27 ? ? ? ? Attitude scales ? ? An attitude is a disposition towards a particular issue, the so called attitudinal object. Attitudinal object may refer to: o A political, economical, or social issue (e. g. , tax on wealth creation, the death penalty etc). o A custom (e. g. , female genital mutilation). o A group (e. g. , UASU, KNUT) The following is an example of an attitude scale: ? An attitudinal scale to test the efficiency of a bus company (e. g. Easy Coach) Strongly agree 1 1 1 1 Agree 2 2 2 2 Agree in some cases 3 3 3 3 Do not agree 4 4 4 4 Strongly disagree 5 5 5 5 Buses are on time Drivers are reckless Fares are too high Service is regular ? Note: researchers as a rule have to compile their own attitude scales to measure the attitudes relevant to their research. Types of attitude scales 1. 2. 3. 4. Likert scale or summated scale Semantic differential The Guttman scale The Thurstone scale These attitude scales are based on different assumptions about the relationship between individuals, their attitudes, and their responses to the items.
The summated or Likert scale ? ? ? This is the most popular type of scale in social sciences. Its popularity stems from the fact that it is easier to compile than any of the other attitude scales (more particularly, Guttman and Thurstone). In Likert scale, subjects have to indicate the degree to which they agree or disagree with its content on, for instance, a five-point scale (e. g. , strongly differ, differ, undecided, agree, strongly agree). Some statements represent a positive attitude whereas others reflect a negative attitude. e. g. , taxing MPs is nothing less than killing them
Taxing MPs is essential for government revenue An attitude scale should contain approximately the same number of positively and negatively formulated items to counteract the acquiescent response style. 28 ? ? The semantic differential ? Each item in a semantic differ