MODERATOR VARIABLES

Moderator variables are characteristics that influence (moderate) the impact of the independent or treatment variable upon the dependent or criterion variable. Moderator variables appear only in studies that include independent and dependent variables - usually experimental studies. In the following hypothesis, for example, the sex of the child is a moderator variable:

  • The use of popular music will increase the appreciation of poetry among elementary school children. This impact will be greater among girls than among boys.

In this example, the use of popular music is the independent variable and appreciation of poetry is the dependent variable. The sex of the child is a moderator variable because the statement indicates that this factor influences the degree to which the independent variable will influence the dependent variable: if the child is a girl, popular music will tend to have one impact with regard to the appreciation of poetry; whereas if the child is a boy, the same independent variable will tend to have a different impact.

In the preceding example, the sex of the child was merely one of many variables that could have been chosen as moderator variables. The following are two other variations of the same hypothesis, with a different moderator variable inserted in each case:

  • The use of popular music will increase the appreciation of poetry among elementary school children. This impact will be greater among older children than among younger children.

  • The use of popular music will increase the appreciation of poetry among elementary school children. This impact will be greater when the teacher is of the same sex as the child than when the teacher is of the opposite sex.
    •

    You can easily see the advantage of using such moderator variables. The researcher (or reader) can discover not only whether the treatment has had the desired impact but also under what circumstances or contexts this impact is likely to vary. This is extremely useful information for helping us generalize the results to other situations, as well as for helping us selectively apply results to specific types of learners within a given educational setting.

    The following are restatements of the first two hypotheses in Review Quiz 2.1. Moderator variables have been added to each hypothesis. To avoid stereotyped language and to show a fuller range of how moderator variables can be introduced into research studies, the wording has been varied in each hypothesis.

    1. Studying Shakespeare leads to an increased appreciation of Western culture among collegebound students but not among non-collegebound students.

    2. Behavior modification will reduce creative behavior among elementary school children in middle-class schools but will increase creative behavior in lower-middle-class schools.

      Now you try to identify the moderator variables in the following hypotheses, again adapted from       Review Quiz 2.1:

    3. Programming in Logo will cause substantial improvements in higher-order thinking skills among slow readers who are not classified as having learning disabilities. However, no such improvements will occur among slow readers who are classified as having learning disabilities.

    4. Delaying reading instruction until the sixth grade will have no adverse impact on reading ability by the time the child reaches adolescence. This absence of impact will be equally true among low-IQ, medium-IQ, and high-IQ students.

    5. Elementary school classrooms that emphasize a language-rich environment will have students who are more eager to share their opinions with their peers. However, in middle school classrooms, language-rich environments will have no such impact.

    The moderator variables in the preceding hypotheses are as follows:

    1. collegebound vs. non-collegebound students
    2. social class of the schools
    3. being classified as having learning disabilities versus not being classified as having learning disabilities
    4. IQ of students
    5. grade level of classroom (elementary versus middle school)

    In each of these examples the original hypothesis has been enriched by the addition of the moderator variable. In each case the moderator variable is a factor that could influence the impact of the independent variable upon the dependent variable. For each example, other factors could have been selected as moderator variables; the actual moderator variables selected in any study depend on what the researcher is interested in learning about the relationship between the independent and dependent variables.

    CONTROL VARIABLES

    Control variables are characteristics that are controlled by the experimenter to reduce any impact they might otherwise have on other variables or on the interpretation of the results of a study. They control extraneous variables (which are discussed later in this chapter). This control can be attained through any of the following methods:

    1. Isolation and elimination. For example, a researcher can control the extraneous effect of intelligence by including only students of average intelligence in the study-excluding low-performing or gifted students. This is an effective strategy described further in the present chapter.

    2. Precise description of the context, treatment, or subjects. A detailed, accurate, and unambiguous description enables persons applying the research findings to make a valid judgment regarding the degree to which the results of a particular study can be generalized to other settings. This is an effective strategy covered in chapters 9 and 15.

    3. Equating across groups. This method is often used in experimental studies. For example, a researcher can control the extraneous effect of intelligence by randomly assigning subjects to the experimental and control groups, thereby making them similar with regard to intelligence (and many other variables). This is an effective strategy described in chapter 11.

    In any research study, the researcher can select and describe control variables to help define the limitations on how far the results of the study can be generalized. Furthermore, in experimental studies, the researcher may select control variables because these factors may otherwise influence the impact of the independent variable on the dependent variable. such extraneous influences would make it difficult to determine the precise nature of the relationship between the independent and dependent variables. Therefore, the researcher controls these extraneous factors to balance or reduce their influence in a particular study.

    Let us examine two of the sample hypotheses that we have been following throughout this chapter, and this time we shall add control variables:

    1. Among high school seniors, studying Shakespeare leads to a greater appreciation of Western culture.

    2. Behavior modification leads to noncreative behavior among elementary school children in the public schools of Gotham City.

    In the first hypothesis, "high school seniors" has been added as a control variable. It is possible that the effect of studying Shakespeare would be different for students of different ages; therefore, the grade level of the students might influence how they would respond to Shakespeare. Thus the researcher has decided to include only high school seniors in the research study. This constraint rules out the extraneous factor, but of course it also restricts the generalizations to those made about high school seniors only. The logic goes like this:

  • Students In the experimental group (who studied Shakespeare) were measured as displaying greater appreciation of Western culture than the students in the control group (who did not study Shakespeare).
  • Since there were only high school seniors in both groups, differences in appreciation of Shakespeare could not possibly have arisen from basic differences between juniors and seniors.

  • Therefore, the influence of grade level (seniors versus juniors) has been controlled; but generalizations from the experiment have also been restricted to those made about high school seniors only.

    • In the second hypothesis, there are three control variables: grade level of children (elementary rather than secondary or middle school), type of school (public rather than private), and location (Gotham City).

    In the preceding examples we added control variables to the original hypotheses before they were enriched by the addition of moderator variables. It is possible (and often desirable) to have both control variables and moderator variables within the same research study. Each of the following examples contains independent, dependent, moderator, and control variables. The moderator variables are indicated by italics, the control variables by boldface.

    3. Among children between the ages of 9 and 11, programming In Logo will cause substantial improvements in higher-order thinking skills among slow readers who are not classified as having learning disabilities. However, no such improvements will occur among 9- to 11-yearold slow readers who are classified as having learning disabilities.

    4. Among children who have no external pressure from their parents to learn to read, delaying reading instruction until the sixth grade will have no adverse impact on reading ability by the time the children reach adolescence. This will be equally true among low-IQ, medium-IQ, and high-IQ students.

    5. In middle-class schools, elementary school classrooms that emphasize a language-rich environment will have students who are more eager to share their opinions with their peers. However, in middle school -classrooms in similar schools, language-rich environments will have no such impact.

    In the third hypothesis, there are two control variables: age of child (9 to 11 years old) and level of reading ability (slow readers are included, but average and fast readers are excluded). There are no 8 year-olds or above-average readers in the study that tests this research hypothesis. The factors of age and level of ability are ruled out as contaminating factors, but the researcher can no longer generalize to students who fall beyond the restrictions set by these control variables.

    You have probably noticed that there is considerable similarity between moderator and control variables. This similarity is a result of both types taking into account factors that may influence the impact of the independent variable upon the dependent variable. The distinction is that moderator variables control this extraneous influence and examine it in such a way as to describe its precise impact. Control variables, on the other hand, merely reduce or describe this extraneous influence without providing any information about its relationship to the independent and dependent variables. Thus, while moderator variables have the potential to restrict, refine, and elaborate generalizations, the role of the control variable is usually confined to merely restricting generalizations.

    Because of the similarity stated in the preceding paragraph, any variable that can be used as a moderator variable in an experimental study can also be used as a control variable, and vice versa. A sensible strategy is to identify as many extraneous variables as possible that might influence the impact of the treatment on the dependent variable; then, some of these can be made into moderator variables and some into control variables, and some can be left uncontrolled. At first thought, it might appear that since the moderator variable does everything the control variable does and more, then we should use the moderator variable as often as possible. This would be good advice, except for an important practical consideration. It is often difficult to obtain the number or variety of subjects that is needed for analyzing a factor as a moderator variable. In addition, having a large number of moderator variables can make a study cumbersome. Similarly, it might at first seem that no extraneous factor should be left uncontrolled. Again, however, there is a problem of feasibility. There are so many factors that could influence the impact of the treatment on the outcome that it is not possible to isolate and control all of them.

    This discussion of control variables has focused on their use in experimental studies. This focus has been useful to point out their relationship to moderator variables. However, control variables are also important in descriptive research. For example, notice the impact of the control variables italicized in the following research questions and hypotheses:

    1. Fifth graders in the public schools of Oregon do better in reading than in mathematics.

    2. A survey of American businessmen showed that they would support higher taxes if they could be assured that the quality of education would actually improve.

    3. The principals who responded to the survey indicated strong opposition to corporal punishment.

    4. The students in the experimental school that used a whole-language approach showed a strong tendency to want to share their opinions with their peers.

    Each of the italicized control variables shows the limitations imposed by the context of the study that will restrict the degree to which the results of each of these descriptive studies can be generalized. If you noticed that many of these control variables were too vague to be of much use, you're right. The usefulness of these control variables would improve with operational definitions (the topic of chapter 4) or qualitative description (the topic of chapter 9).

    INTERVENING VARIABLES

    Intervening variables are hypothetical concepts assumed to be created by the treatment and to have an impact on the observable outcome. Treatments produce intervening variables (which are invisible, perhaps inside the brain of the learner), and intervening variables then produce observable outcomes. This variable intervenes in the sense that the treatment does not produce the observable outcome directly but rather through the mediation (intervention) of this invisible, conceptual, hypothetical, internalized process. The intervening variable is the basic subject matter of theoretical research (discussed in chapter 17).

    To take a brief example, a researcher might find that computer simulations cause students to do well on science tests. In fact, the computer simulations themselves probably do not cause the improved performance on the tests. The simulations probably induce some internal change within the students (such as improved motivation, greater attention to the task, perceived relevance, or something else), and this internal change is what eventually leads to the difference in performance. Determining and examining intervening variables helps us understand educational principles better. In addition, such variables can be useful for practical reasons: we may be able to find less expensive ways to bring about these intervening variables than, for example, relatively expensive computer simulations.

    In the following examples the hypothesis is stated (without the moderator variables or control variables), and then a possible intervening variable is stated in parentheses:

    1. Studying Shakespeare leads to a greater appreciation of Western culture. (perception of important similarities within Elizabethan and contemporary Western culture.

    2. Behavior modification leads to noncreative behavior among elementary school children. (an urge to conform in order to be reinforced as efficiently as possible)

      Now you try it. Identify possible intervening variables for the following hypotheses:

    3. Because Logo offers opportunities for improved planning ability arising out of the involvement in microworlds, programming in Logo will cause improvements in higher-order thinking.

    4. Delaying reading instruction until the sixth grade will have no adverse impact on reading ability by the time the child reaches adolescence.

    5. Language-rich environments will make children more eager to share their opinions with their peers.

    Here are possible intervening variables for the preceding hypotheses:

    1. Background information that makes it easier to understand aspects of Western culture
    2. A tendency to refrain from doing things that do not gain extrinsic reinforcement
    3. Improved planning ability arising out of the involvement in microworlds offered by the programming language
    4. Self-motivation, which replaces external requirements
    5. In this case, none is stated. A reader or researcher could construct one through such processes as reasoning about the evidence arising from the study and reflecting on theoretical information from the research literature. Through such reasoning, stimulation of language and social processes might emerge as the intervening variable.

    In these examples, the intervening variable has been appended to the barest possible statement of the hypothesis. This has been done for the sake of simplicity. In actual practice, the moderator variable should take into account all the variables in the research hypothesis.

    Unlike the other variables, the intervening variable is usually not stated as part of the hypothesis. Rather, it is stated at the culmination of the review of the literature (prior to- the hypothesis) as the specific rationale behind why the hypothesis is going to be stated in the form it will take.

    EXTRANEOUS VARIABLES

    An extraneous variable is a factor that produces an uncontrolled, unpredictable impact upon the dependent variable. Extraneous variables weaken research studies because they introduce ambiguity into the research process. To the extent that they are uncontrolled, they render uncertain the conclusions we can draw from a study. The threats to internal and external validity, discussed in chapters 10 and 1S, are extraneous variables, and the whole purpose of the strategies discussed in those chapters is to minimize the impact of such extraneous variables. To the extent that extraneous variables can be brought under control, the results of a study are strengthened. One good reason for replicating research (discussed later in this chapter) is to reduce the possibility that conclusions will be weakened by extraneous factors.

    THE RESEARCH VARIABLES COMBINED

    Table 2.5 summarizes the major characteristics of each of the research variables. Our discussion of research variables has focused largely on experimental research. This is because only in experimental studies are we likely to find all the research variables in a single hypothesis. We may have given the false impression that most educational research is experimental. Actually, only about 10% of published educational research is experimental; the rest is either descriptive, qualitative, or correlational. Table 2.1 briefly described these various types of research and indicated where each is discussed in this book. Also, the categories in that table are not mutually exclusive; for example, a quantitative or experimental study may also employ some qualitative methods, and a meta-analysis often combines the results of several experimental studies.

    Most studies do not contain every type of variable. Whatever variables emerge from the analysis of a problem setting should be clearly identified and appropriately stated in a succinct research question or hypothesis. Subsequent chapters will describe how to operationally define these variables and incorporate them into effective research studies.

    Table 2.5 Summary of Research Variables

    Variable

    Definition

    Mnemonic

    Other Terms

    Dependent

    The outcome that is expected to result from a treatment

    Such variables are dependent in the sense that they depend on the treatment

    Outcome variable

    Effect

    Result

    Criterion variable (in prediction studies)

    Independent

    The treatment or condition that is expected to produce an outcome

    The independent variable is independent in the sense that it does not depend on the outcome variable

    Treatment

    Experimental procedure

    Cause

    Predictor variable (in correlational studies)

    Moderator

    Characteristic that influences the impact of the independent variable upon the dependent variable

    The word moderate can mean to modify, as one weather pattern may moderate the impact of another (if that doesn't help, modify and moderate both begin with mod)

    Interaction

    Interactive factor

    Control

    Characteristic that is controlled by the experimenter in order to reduce any impact this factor might other wise have on the interpretation of the results of a study

    The whole purpose of this variable is to control extraneous influences

    Context variable

    Limitation

    Restriction

    Intervening

    The hypothetical factor that is created by the independent variable and has an impact on the dependent variable

    This variable intervenes in the sense that it occurs during the time that intervenes between the cause and the effect

    Underlying cause

    Psychological construct

    Extraneous

    Factors that produce uncontrolled, unpredictable impacts upon the dependent variable

    This variable is an extra factor, something that is extraneous to the current study and therefore may cause confusion, unless it is controlled

    Contaminating factor

    Threat to internal or external validity

    REVIEW QUIZ 2.2

    1. Examine the following research hypothesis and identify each of the research variables requested below:

    Formal operational high school biology students will use problem-solving heuristics to solve genetics problems, but concrete operational high school biology students will use a rote-memory strategy to solve the same kind of problems.

    Identify the following variables:

    Dependent

    Control

    2. Examine the following research question and identify each of the research variables requested below:

    How do K-3 students who use the whole-language method to study reading differ in their comprehension strategies from those who use a phonetic approach?

    Identify the following variables:

    Independent

    Dependent

    Moderator

    Control

    3. Examine the following research hypothesis and identify each of the research variables requested below:

    High school students who study Latin for two years will develop better English vocabulary skills than those who do not study Latin. This difference will occur both among those who are in advanced-placement English classes and among those who are in regular English classes. This difference will occur because of an increased ability to break a word down into its component parts.

    Identify the following variables:

    Independent

    Dependent

    Moderator

    Control

    Intervening

    Click here for answers.

    Source: http://education.calumet.purdue.edu/vockell/research/chapter2.htm
    Last modified: Wednesday, 14 March 2012, 10:28 AM