What is Experimental Research ?
Experimental research is a scientific investigation that involves manipulating and controlling independent variables to observe their effect on dependent variables, allowing researchers to establish cause-and-effect relationships.
Concept of Cause
In order to establish a cause and effect relationship, the primary principle is to prove that in an experiment first the cause happens and then the effect follows it. Four kinds of relationships may occur between cause and effects, they are as follows:
- Either the cause can occur or it can be made to occur and its effect can follow immediately.
- Either the cause can occur or it can be made to occur but its effect cannot follow immediately.
- The cause cannot occur but it can be made to occur, however, its effect can follow immediately.
- The cause cannot occur as well as it also cannot be made to occur and its effect cannot follow immediately.
The absolute motive of a research is to establish a cause and effect relationship. This is the reason why experiments are conducted in such a way that only those causes are considered which can establish a logical relationship with the effects consistently.
Experimental research design or "design of experiments" or "causal research design" studies causality, i.e., cause-and-effect relationships between Research variables. In this type of research design, researcher selects two similar groups from the target population, one is called 'experimental group' and the other is 'control group'. While control group is kept constant and free from any interference. experimental group is exposed to set of experiments. At the end of experiment, the changes in experimental group are measured by comparing it from control group. Therefore, the factors affecting the change are identified.
Experimental researches are controlled and structured in nature. To understand the causal relationship, it is necessary to manipulate one or more independent variables to measure the effects on dependent variables. Therefore, experimental or causal researches help in collecting experimental knowledge or information, which is based on experimental data rather than theories. This research may be helpful in understanding an event or enhance the performance in . a particular field. Experimental research is important to society as well for business, as it helps in predicting the future activities. It provides an outline of activities to be performed to test the relationship between various variables by proving hypotheses. In other words, it is a structured design that allows the researchers to carry-out experiments.
Causal Relationships
In case of explanatory social research, when two or more variables establish a causal relationship among them, then it justifies the topic on which research is being conducted. John Stuart Mill referred a cause us a notion of constant conjunction. In other words, any event or entity can be a cause of any other event or entity only if both the events are taking place at the same time; provided that, the cause should take place before the effect and other probable causes should be ignored to keep the relationship authentic. Various quantitative researchers have used causality as a separate domain of research by using experimental methods, non experimental methods, or structural equation modelling along with the non-experimental data.
Types of Causal Relationships
The types of causal relationships are as follows :
1) Direct Causal Relationship:
When an entity of event is believed to be a direct cause of occurrence of any other event or entity, then that relationship is known as direct causal relationship.
For example, inclination of an adolescent child toward drugs depends on his relationship with his parents, i.e., poor relationship is believed to be a cause of drug addiction.
2) Indirect Causal Relationship :
When a variable impacts another variable, which in turn impacts one more variable, then the relationship between the first and the third variable can be referred to as indirect causal relationship or the mediated relationship. In such cases the second variable is known as intermediary variable, mediating variable, or mediator.
For example, a person who fails to achieve his pre-determined goal may, get frustrated. Due to this frustration, he may behave aggressively with any other person. Hence, in the stated example, the failure of achieving the goal shares an indirect causal relationship with the aggressive behavior of the person. And the frustration aroused due to the failure is the intermediary variable. It is very likely that according to one theory a relationship is direct and according to another the same relationship is indirect.
3) Spurious Relationship :
The relationship, in which the two variables are related not because of the occurrence of each other but because of the occurrence of a common cause, is known as spurious relationship.
For example, a research is conducted in order to identify the relation between feet size and the verbal ability of human beings, and it is found that bigger the feet a person has, more is his verbal ability. This is certainly not an example of a causal relationship instead they are correlated because both of them have a common cause, i.e., both of these variables are dependent on the age factor. Infants have small feet and they cannot speak clearly, as they grow young they learn to speak a little bit, and as they grow old they become more efficient in verbal communication. The verbal ability and the feet size are correlated to each other because of a common cause; hence the relationship among them is said to be spurious.
4) Moderated Causal Relationship :
Similar to indirect causal and spurious relationship. moderated causal relationship also involves minimum of three variables. In this kind of relationship two variables share a causal relationship but it depends on the value of third variable involved in it.
For example, certain treatment gives relief to the stomach ache, but only for male patients and not for female patients. Hence, the causal relationship among that treatment and the relief in stomach ache depends upon the gender of the patient. Gender, in this case, can be termed as the moderator variable as the relationship between two variables entirely depends on it.
5) Bidirectional or Reciprocal Causal Relationship :
In case of bidirectional relationships, both the variables are related to each other on the basis of certain reasons.
For example, in order to prevent the pregnancy a method is used named as rhythm method. Here, the bidirectional relationship can be seen among the belief of women regarding the effectiveness of method and their attitude towards the method. Some women believe that this method is effective; therefore they may develop a positive attitude towards the method. In the same way, some other women have a positive attitude towards the method; therefore they may think that the method is effective.
6) Unanalysed Relationship :
Unanalysed relationship is that relationship where the two variables are correlated with each other but the reason behind their correlation is neither evident nor the researchers and theorists want to state them.
Conditions for Causality
There are three basic conditions that are required for a relationship to be causal. Although these conditions alone are not sufficient to prove a cause-and-effect relationship, but these are necessary. Following three conditions must be satisfied for the causality in / experimental research design :
1) Concomitant Variation :
Concomitant variation is the degree to which the causing factor (X) and effective variable (Y) show the relation as predicted by the hypothesis. The evidence of relationship can be measured in qualitative or quantitative manner.
For example, in qualitative case, to estimate the relationship between the performances of team members (effect) and their salary (cause) can be estimated by formulating a hypothesis supporting the concomitant relation. Therefore, this concomitant relation would show that increasing salary increases the performances of team members, while reducing salary decreases their performances.
2) Time Order of Occurrence of Variables :
According to this criterion, the causing event can not happen after the effect has been taken place. It must occur before or concurrently with the effect. An effect cannot be generated by the event that occurs after the occurrence of effect. Though, it is possible for an event to act both as cause and an event in same relationship.
For example, team members who get higher salary (cause) perform better (effect), while the employees who perform better (cause) can get higher salary (effect).
3) Elimination of other Possible Causal Factors :
The last condition for the causality in the experimental research is the elimination of other possible causal factors. As per this criterion, there should be only one causing factor keeping other factors constant.
For example, the salary of the team members can be the only possible causing factor, when all the other variables such as working environment, employee involvement, recognition, promotion, etc. are kept constant while estimating the relationship. While on one hand, in an after the-fact examination, controlling all the variables is not possible, on the other hand, controlling some of the causing variables is possible up to some extent. The effects of several uncontrolled factors may be stabilized for determining only the random variations from them.
Basic Principles of Experimental Design
Experimental designs follow three basic principles, which are as follows :
1) Randomization :
Treatments are assigned to the experimental units with the help of the process of randomization. This process denotes that the possibility of assigning all the treatments is equal, The experiment material is divided into several small units, which are known as experimental units and the conditions of experiment whose effect is to be measured are known as treatments. Randomization is done with the intent to remove the element of biasness and the possibilities of unnecessary and uncontrollable variations. Randomization along with the replication is also beneficial because it serves as a basis on which a valid statistical test can be conducted. Well shuffled pack of cards, well-shaken box containing balls, dice, etc., can be used for the process of randomization.
2) Replication :
The literal meaning of replication is repetition, i.e., in this process the basic experiment is repeated. Basically, during replication all the treatments are tested in the actual experiment. Due to the fact that all the experimental units cannot be identical, slight variation can be seen in every experiment. In order to get rid of such variations, various experimental units should be used. Hence, researchers prefer to conduct the experiment several times. One repetition is referred to as replicate and experimental material derives the number of replicates, their shape, and their size. Purposes behind using replication are as follows :
- Accurate estimation of experimental errors by recording the variation so that this variation can be compared with the other variations when same treatment is applied again on the same experimental units.
- Reduction of experimental errors in order to increase the accuracy or precision of experiments. Precision is tool for measuring the variations of the experimental errors.
3) Local Control :
When randomization as well as replication fail to remove all the unnecessary variations, then arises the need of improvement in the technique of experiment, i.e., the researcher needs to re-design the experiment, keeping in mind that the unnecessary variations. are to be kept under control. In order to help the researchers in the stated problem, the principle of local control can be used. Local control refers to certain extent of balancing. blocking, and grouping of the experimental units.
'Balancing' means allotment of treatments to an experimental design in order to keep the arrangement of treatments balanced. "Blocking' refers to the process of collecting all the units together in order to form a group of comparatively units. The block so formed is also known as replicate. One of the prime objectives of local control is to reduce the errors in experiments so that the efficiency of experimental designs can be improved. It is very important to understand that the local control is. not similar to the term 'control', because in the field of experimental design control refers to the treatment.
Treatment and Control Group in Experimental Research
The notion of control establishes a fact that two persons, groups, or conditions are not different from each other. The thing that differs among them is their interest. In such cases, the research is said to be valid internally and the method of difference is also applicable.
For example, two similar groups of persons are selected. First group is tested by giving them certain treatment; hence that group is treatment or experimental group. Whereas, no such treatment is given to the second group; hence that group is control group. In such experiments, control group is treated as the benchmark and the experiment group is compared with it because initially both the groups were similar. Therefore, the changes due to experiment can be easily seen among the members of experiment group and the credit of these changes can be given to the treatment.
In case of experiments, the experimenter intently introduces certain factor in one group and keeps the other group away from it. Then the experimenter observes both the groups in order to spot any change in one group. If any change occurs, then that change is certainly because of the factor introduced by the experimenter in one group.
Some changes take place with time and these changes are equal for everyone. But, if the groups are similar to each other, some changes will also take place in both of them. This is the reason why the researchers use experiments and not the time factor, for identifying the changes. Because, with the help of experiments it can easily be concluded that the changes occurred to the treatment group are due to the factor induced in that group.
Control groups are formulated either by allocating random experimental units from outside or by the experimental units themselves. In any research, the experimental units are selected out of their population on a random basis and then out of the selected units, they are divided into the experimental groups and the control groups. Control groups are also known as contrast groups or baseline groups.
In order to arrive at the inference, the outcomes of both the groups are compared and influence of the treatment is evaluated by measuring the differences between the outcomes of control group and different experiment groups. In case of true-experimental studies, the experimental units are selected from their population on a random basis, but their allotment in the experiment and control groups is based on the baseline data or control data. Then this control data is compared with the assumed outcome data, which makes the experimental data their own control group. On the contrary, in the case of quasi-experimental studies, researchers do not assess the effect of treatments on the experimental units.
Types of Experimental Research Design
According to the Campbell and Stanley, experimental research design can be described with the help of following symbols :
X = A set of treatments introduced on experimental group.O = Observations or measurements taken on independent variable. (In case of more than one observation, these can be denoted with the help of subscripts, i.e., O₁,O₂...).R = Random assignment of the test units.
Experimental research designs can be carried out using following methods :
Pre-Experimental Designs
In pre-experimental designs, treatments are not allocated to the subjects randomly. Therefore, these designs cannot be called true-experimental designs, as these techniques do not deal with the challenges which have occurred due to loss of internal validity. These designs are suitable for the cases, when it is the only possible solution for an experiment. Various pre-experimental designs are as follows :
1) One-Shot Case Study :
This technique is also known as 'after-only design'. In this technique, only one test unit is assigned a treatment X, and then a single measurement O, is taken on dependent variable. The test units are not randomly assigned to the treatments, and the test units are selected by the researcher.
For example, if a primary school teacher wants to see if praising children cause them to become more confident. He tests it with two students of second standard, and praises them. He finds that they are more confident.
While one-shot case study is the most basic form of experiment, it has its own limitations. It does not all the researcher to compare the measurement O₁, with the dependent variable before the treatment. Along with this, the measurement of one-shot case study can be affected by many extraneous variables, which can be controlled. This in turn affects the validity of the measurement. Therefore, this type of research suitable for exploratory research rather than experimental research. This technique is opted only when it the only solution to the problem.
2) One Group Pre-Test - Post-Test :
In one group pre-test-post-test design, only one group is involved which exposed to the treatment. Here, the researcher measures the subjects before and after the implementation treatment. It is the improvement over one-shot case study, as it allows researchers to compare the change in objects from before and after stages.
For example, a sales manager may wish to conduct a training programme to enhance the knowledge of sales team members. The sales manager may measure t knowledge of team members as 'O₁'. As the training programme is completed, the sales manager may again measure the knowledge level of team members as 'O₂'.
This technique is widely used in marketing research. It also suffers from disadvantages. There is possibility that other extraneous variables cause the change rather than the treatment.
3) Static Group :
In this technique, there are two groups upon which measurements are to be taken. The first group is experimental group which is subjected to the treatment, and the other one is control group, which is not subjected to the treatment. Here, the measurement of experimental group is taken after being exposed to treatment, while the control group is measured without treatment. The outcome of the experiment is measured by comparing both the measurements.
For example, a researcher may wish to compare the effect of a medicine by giving it to a patient, and comparing his condition with another similar patient without the medicine.
Although, this technique is more reliable than previous techniques, but it also has some disadvantages. This technique does not assure that both the control group and the experimental group are equal in every aspect If the groups are selected randomly, and the group elements are not similar in nature, then the outcome ma differ which may affect the validity of findings. When the group elements are selected by the researcher then it is called as an after-only design with control group.
True Experimental Designs
True experimental research design is considered to be the most accurate type of experimental research design. In this type of experimental designs, researchers try to conduct research to prove some hypotheses using statistical techniques. True experimental designs are as follows :
1) Pre-Test - Post-Test Control Group Design :
In this technique, a measurement is taken both from the experimental and control group before the treatment is administered on control group. After the treatment is subjected, a post-test is conducted on both the groups to measure the changes in the groups. The difference between the measurements can be calculated statistically. The criterion for conducting this technique is that both the groups should be similar in every possible aspect. The basic philosophy behind this design is that both the groups would be equally affected by the presence of any extraneous variable.
For example, this technique can be used to measure the effectiveness of an advertisement. For this, two groups will be randomly selected, named 'experimental group' and 'control group' respectively. A questionnaire will be given to them to measure their perspective regarding the product. After that members of the experimental group will be shown the advertisement. After showing the advertisement, the measurement will be taken from both groups to see the changes in perspective regarding the product.
2) Post-Test Only Control Group Design :
In post-test Only control group design, experimental and control group are selected from the target population, which are identical nature. Before the introduction of the treatment, no measurement is taken from any group. Only after the introduction of the treatment, the level of phenomenon is measured in both the control and experimental groups. The resulting effect of the variable may be calculated by subtracting the control group level from the experimental group level. The above example can be implemented through this technique also. The difference would be that researchers would not test the perspective of group of people before showing the advertisement related to them. The perspective of the people would be measured right after showing the advertisement.
3) Solomon Four Group Design :
Solomon four-group is an improvement over pre-test - post-test design., This design introduces two. additional control groups, which helps the researchers to assess the influence of confounding variables on the measurement. It also allows estimating the changes caused due to pretest on the subjects. In this experimental design, different combinations of tests are conducted to identify the extraneous and confounding variables and reduce their effects on the outcome. Although this design is beneficial from the perspective of accuracy, but executing this design incurs huge cost and takes a lot of researcher's time.
For example, 100 teachers are selected and randomly divided into four groups of 25 teachers and named 'experimental group-1', 'experimental group-2', 'control group-1', and 'control group-2 respectively. Now, first experimental group would be given a faculty morale questionnaire and would receive the treatment in form of sensitivity training. The control would be given the questionnaire and would not receive any treatment. On the other hand, second experimental group would receive sensitivity training and the second control group will directly be post-tested in form of questionnaire. As soon as the groups are pretested and receive treatments, they would be again given the questionnaire, Now, the effects on various groups of teachers can be measured by comparing the groups.
Quasi-Experimental Designs
Quasi-experimental research designs are applied when true experimental designs cannot be applied. These techniques are easy and economical in nature. These research designs are suitable for following conditions :
- Researchers have complete control over the measurements.
- Researchers cannot control the projection of treatments and randomization of test units is not possible.
Some of the major quasi-experimental designs are as follows :
1) Time-Series Design :
In this design, the researcher takes a series of periodic measurements on the dependent variable. The treatment is introduced either manually by the researcher or occurs naturally. As soon as the treatment occurs, the periodic effects in measurements are estimated.
By measuring the dependent variable before and after the treatment, researcher becomes able to identify and control the extraneous variables: Periodic assessment not only affects the immediate measurement to the treatment, but also affects all the measurements. By applying randomization in selection of test units biasness in research outcomes can be reduced.
For example, the death rate of people due to Ebola disease, can be measured periodically, i.e., before and after the medication of people in successive years to measure the effects.
While the people suffering from this disease can be measured before the treatment, the people can also be observed after introducing the treatment in form of medication. This will enable the researchers to see the effect of medication on people.
2) Multiple Time-Series Design :
This design is slightly different from the time series design in a way that it includes a control group for measurement. The efficiency of this design depends on the fact that the effect of treatment needs to be measured twice, i.e., against the measurements of experimental group before the treatment, and against the control group as well. Another consideration while performing this research is to select the test units of control group carefully. Taking the above example, the same observation can be made with the help of a control group. The members of control group would not receive the medication. This technique helps in drawing the conclusions in a better way.
Statistical Designs
The most common statistical designs are as follows :
1) Randomized Block Design :
Randomized block design has evolved from agricultural research, where the researcher introduces various treatments to different blocks of land, so that their effects on the yield of crop can be assessed. Though there may be difference in the characteristics of land due to which the yield of crop can be affected. To identify the differential factors, the researchers introduce the treatments to the plots in each block randomly. The number of plots in a block is equal to the number of treatments, so that plots from each block can be selected for different treatment. As soon as the treatment is introduced, the production of crop is measured by using statistical techniques to analyze the effect of treatment.
2) Latin Squares Design :
In this research technique, the researcher aims to control the variation in two factors. The design forms a square as there is equal number of rows and columns. This technique is adopted to identify the extraneous variable causing the change. In this research, all the possible combinations of these two variables can be estimated multiple times. This design is used to reduce the effect of nuisance factors. An important criterion for conducting this research is that the number of many test units should be equal to the number of treatments.
3) Factorial Design :
In factorial experimental design the researcher tests two or more variables at the same time. It tries to find out whether the two variables combine to form the observed response or they combine independently. This technique is suitable when there are three or more experimental variables, and the test units are selected randomly. The major disadvantage of this method is that it' involves complex calculations.
Uses of Experimental Research Designs
Experimental research designs have the following uses :
1) Understanding Consumer Behavior :
By using the experimental research designs, the researchers observe the current behavior of consumers. With the help of current behavior, they try to predict the future behavior and to reveal the past behavior as well. The psychologists of experimental research field are aware of very wide range of phenomena such as learning, personality, cognitive development, etc.
2) In the Field of Sales :
Researchers use the methods of experimental research in order to evaluate the effectiveness of the promotional efforts of the company. It also helps in identifying the affect that the advertisements made on the consumer buying behavior.
3) Searching for Facts :
These designs are used by the researchers to extract the research related Information. In this, the variables of the experiment are used to thoroughly understand the problem under inspection.
4) Testing Validity of Empirical Proposition :
Researchers use these research designs in order to validate the assumptions of a theory and to compare the actual outcome with the intended outcome. This testing assists the researchers in measuring the consistency and authenticity of research.
5) Business Communication :
Business communication is a very crucial aspect of a business and therefore, it is very important to make the business communication effective. Some of the aspects of business communication, if studied with the help of experimental research designs, may provide valuable output and information to the marketers.
Importance of Experimental Research Design
The significance of experimental research can be understood by following points :
1) Cause and Effect :
The major advantage of experimental research design is that the cause of a particular event can be identified here. Other research designs either explain the event or describe the population, but do not determine the cause behind an effect. By conducting an experiment using random assignments, and participants unknown to the causing factors, would enable the researcher to observe any deviation due to the experiment.
2) Reliable Outcomes :
Another benefit of experimental research is that the outcomes of experimental researches are highly reliable, as the procedure is conducted in a controlled environment using quantitative measures, and random assignments. This distinctive feature allows the researchers to generalize the findings as well as replicate the experiment with similar sample drawn from the same larger population. The samples drawn are true representative of the population.
3) Provides Helpful Insight :
These designs are very useful in providing important insight for solving immediate problem.
For example, by researching on different techniques of motivation, a suitable and effective motivation technique may be developed for the employees of an organization.
4) Control over Variables :
In experimental research design, the researcher can control different variables in the environment. Thus the researcher is able to determine the individual impact of all possible variables. Moreover, the interrelationship among variables can also be measured more effectively.
Limitations of Experimental Research Design
Experimental research design has following limitations :
1) Artificiality :
The experiments are performed in artificial settings which lack real-life conditions. Such unnatural environment disturbs the genuine behavior of respondents, hence producing a false picture. As these experiments are entirely different from the real-life environment, the outcomes of such experiments may not be used for other live situations.
2) Biasness by Researchers :
Since in experiments researchers input the causing variables manually, therefore it is susceptible to "selection bias" on the part of the researcher. Many times because of convenience in execution, the researcher can manipulate the inputs. As a result, the outcomes are not completely reliable and valid. Moreover, the easiest and cost-effective sample is selected by the researcher for the study and findings of this study are generalized for further use.
3) Modified Responses :
The responses of the participants may be influenced by many factors in the surrounding. Instead of giving genuine responses, the respondents may respondent what the researcher wants to hear or may modify the response as per the subject of research.
4) Impossible to Control All the Variables :
The much known feature of experimental design is the control over different variables present in the environment. However, all the factors cannot be entirely controlled as it is not possible to identify all the potential variables affecting the experiment. Thus, it is very difficult for the researcher to have complete control over the thinking and behavior of the respondents.
5) Uncertainty of Actual Responses :
The responses given by the respondents in such designs are uncertain as it is difficult to differentiate between the true and manipulated responses.
Variables in Research
Those factors and elements that show some variations at different point of time and in different situations are known as variables. Variables also have the ability to impact the outcome of any study. Nonetheless, they are crucial for any study because with the help of variables a researcher is able to identify the differences.
As the name suggests, variables have minimum two values. Some of the variables are tangible and visible in nature like weight, height, gender, etc., whereas others are very descriptive such as, IQ, perception, beliefs, etiquette's, self-esteem, etc. Hence, it can be said that variables are those facets which one person has different from the others.
In the process of designing of quantitative research, one of the important steps is to identify the variables associated with the topic of research. On the other hand, qualitative researchers do not give as much consideration to the variables as their quantitative counterparts give, yet they also have to outline the phenomena they are going to do research on. The most important categories of variables are discussed in detail. Though they are explained under the heads of quantitative research methods, the qualitative research also comprises of similar examples.
Independent and Dependent Variables
Variables share mainly two types of relationship among each other, they are as follows :
1) Independent Variable (IV) :
Independent variables are those variables which vary on their own but they also affect the other variable(s) under research. It is very likely that a variable which is independent for a particular study may be a variable under control for any other study. Similarly, in certain study the independent variables may slightly be correlated, then in such cases they are not said to be independent variable.
For example, the relationship between crop yield and rainfall is such that the rainfall is an independent variable, whereas the crop yield depends on it. But, it is not necessary that the rainfall will act as an independent variable in all the situations.
2) Dependent Variable (DV) :
Dependent variables generally get affected by the changes of independent variables. As these variables change frequently, the need of measurement, monitoring. and prediction of these variables arise. Dependent variables are also known as criterion variable.
Concomitant Variable
Those variables whose values are generally available before the commencement of research and they are also correlated with the outcome of the research, are known as concomitant variables. Also, these variables are not one of the reasons due to which the research is being conducted. The examples of such variables are covariates and blocks. How to include these variables in the research design is one of the major challenges of research design process. Strategies used to overcome this challenge are as follows :
- Make use of concomitant variables as covariates.
- Consider the concomitant variables as blocks and use them as the subject of research.
- Provide certain range to the values of concomitant variables so that the range intervals can be used as blocks. For example, for the respondents of different age groups, divide the groups in the intervals of 20 years.
The concomitant variables cause the undesired variation among the other dependent and independent variables. This characteristic of the concomitant variables reduces the risk of biasness. On the other hand, these variables give rise to the error variance, which in turn reduces the chances of recognizing the actual differences. To overcome these kinds of demerits, an ideal research design strategy is required which can reduce or nullify the effect of undesired variations caused by the concomitant variables. There exist two such strategies, which are as follows :
- Experimental control
- Statistical control
Moderating Variables
Every relationship comprises of minimum one dependent and independent variable. The second important independent variable under consideration is known as moderating variable. Such variables are involved because of the belief that it can make an important contribution towards the relationship of dependent and independent variables.
For example, the incentive-based compensation structure leads to improved performance of the employees and this effect is more evident among the young employees. Hence, in the stated examples, incentive-based compensation structure is the independent variable and performance of employees is the dependent variable, whereas improvement of performance mainly seen in the young employees is the moderating variable.
Extraneous Variables
In any situation, the extraneous variables exist in a large number and it is very likely that they influence the relationship under consideration. But, the effect of these variables is very less, and therefore most of them can be ignored. And the remaining, which might be of some importance, are either considered as independent or as moderating variables.
For example, issuing of discount coupon on the purchase of raw cereals will not heavily affect the sale because of numerous reasons such as, people may think that the government is about to increase the tax rates, election is scheduled, heavy rain is forecasted, etc.
Intervening Variables
According to Tuckman, The intervening variable is that factor which theoretically affects the observed phenomenon but cannot be seen, measured, or manipulated; its effect must be inferred from the effects of the independent and moderator variables on the observed phenomenon".
In connection with the above mentioned example of incentive-based compensation structure, some employees can perceive that their overall salary is increased and some may derive job satisfaction from it. These variables can be termed as intervening variables.