The usefulness of a study or research is usually dependent on the accuracy and level of confidence of the findings. This will be subject to the procedures and steps as well as the variables used when carrying out the research. The level of accuracy and confidence of a research is determined by its validity; which is the level with which correct inferences are made from the research study results. A researcher always considers the confidence that all observations are from the given variables and the research was not manipulated. Such a kind of confidence shows the internal validity. Researchers use outcomes of their studies to make a given claim about the sample and about the whole population of study. The ability to make a claim or a conclusion or the population of study, which is called a generalization, is dependent on the research’s external validity.
Internal validity
Internal validity is simply described as how efficiently comprehensively and accurately a research was conducted. Furthermore, it shows how confidently an individual can make a conclusion that the observations from the research independent were fully derived from the independent variables and no other external or unrelated variables. Internal validity is used in experimental research to determine whether treatment causes a difference between subjects in experimental and control groups. It is used in descriptive studies such as correlational studies to find the quality or accuracy of the study in question. A strong level of internal validity implies that there are reliable measures for both the dependent and independent variables and justification linking the dependent and independent variables. It also helps to do away with extraneous variables. Internal validity deals basically with causal control.
An experiment is said to be internally valid when it clearly shows the cause and effect relationship that exists between dependent and independent variables. For example, an experimenter can observe that there is an increased probability of death due car accidents. This will lead to the conclusion that car accidents cause death. For this particular conclusion to be considered internally valid, the experiment has to be designed other conditions apart from car accidents are not considered as potential causes of death.
Threats to internal validity
Threats of internal validity are the extraneous variables that, if not regulated or controlled, can jeopardize the internal validity of the experiment. These are variables or factors that are related to the variables in the study but should not be considered when carrying out the research.
Maturation
It is the natural and gradual change of a subject over a long period of time. For instance, when a study takes a long time the subjects could grow weary and lose the motivation due to hunger or thirst or old age. Basically, the performance of subjects is dependent on the dependent variable rather than the independent variable. They are simply old or less motivated.
Selection
This is the nonequivalent groups’ effect on the validity of a research. The control and experimental groups must be equivalent functionally at the start of the study. If the two subjects are equivalent, the difference observed between the two groups as indicated by the dependent variable’s performance could be because of the independent variable and not organismic variables. If the groups being compared are not identical at the beginning of the study, the effects observed could be due to such differences.
History
It is the effect caused by external events of the past on subjects between various measurements taken in a given experiment. The history leads to unplanned or extra independent variables. Generally, such experiences vary from subject to subject hence; there are different effects on the responses of different subjects. Research studies which take repeated measures about subjects over a period of time are highly likely to be interfered by historical variables compared to those whose data collection is within a short time without repeated measures.
Testing
It is the manner in which a pretest affects the performance of subjects in the post test. Most experiments pretest subjects so as to establish that every subject begins the experiment at a common level. The effect of pretesting programs is that it can jeopardize the performance of the subjects on later tests, such as those used to act as dependent variables, which measure a common domain.
Statistical regression
This is the phenomenon whereby retested results have the tendency to regress to the value of the mean. If subjects in a given study are selected on participant basis because of the extremely low or high performance measures, a retest of all subjects will produce a totally different score distribution and the mean for the new distribution will be very close to the population mean. For instance, if all selected subjects have initial high scores, the group’s mean retest is usually lower than the original mean. Similarly, if the group mean was initially low, the retested mean will be higher.
Instrumentation
Instrumentation is the reliability, objectivity and validity of the study measurements. Biased information is usually unreliable and is a threat to the study’s internal validity. Furthermore altering the methods of taking measurement during the study is likely to affect the subjects’ measurements.
Selection Interactions
This is the method of selection relates with other factors such as history, maturation or instrumentation. These kinds of interactions are likely to cause bias.
Attrition
It is also called experimental immortality. It is the potential occurrence of bias that is dependent on which subjects drops out or stays in the research. If one group of comparison shows a high attrition level than others, then the differences observed are questionable. If the observed differences were given by different rates of dropping out or by an independent variable, it results in attrition. The other threat related to attrition is mortality since it has a similar effect to drop outs.
Minimization of internal validity threats
For the researcher to obtain a strong internal validity, he or she must minimize the threats. This can be achieved by ensuring that the conditions for the research are standardized so as to reduce threats resulting from instrumentation and historical factors. Comprehensive and accurate information concerning the subjects is needed to reduce threats from selection and mortality factors. The procedures of the research should be adhered to the later to avoid confusion and instrumentation problems. Finally, the choice of an appropriate research design plays a significant role in controlling internal validity threats.
External Validity
This is the extent to which a research’s findings can be used to make a generalization or conclusion concerning a whole population, settings or people. Therefore, it addresses the capability to use study findings to make generalizations. To achieve strong external validity, the researcher needs to undertake a probability sample of respondents or subjects obtained randomly from a well defined population. Using the different samples from a given population, obtain the measurements of each sample then find a mean. With a strong external validity, the conclusions will be made confidently. This is exactly what happens with public opinion studies.
External Validity Threats
This is a show of how the conclusions made using research findings can be wrong. For example, a conclusion can be generalized but in a different context. External validity has three major threats which are places, people and times. However, there are other threats.
Place or Setting
Subjects’ performance in a research can be dependent on their reaction to the experiment’s setting rather than to the independent variable. The outcomes of a study could also vary from one place to another. This implies that if the study is carried out at a different place from the initial one, the outcomes will be different even if the same apparatus, subjects or measurements are used.
Pretesting
Subjects or respondents in a study could react differently in subsequent tests if they are subjected to pretests. For such an instance, the researcher has no basis to make a conclusion that the part of the population that was not pretested would produce a similar performance to those who were pretested in the study. The researcher can, however, restate the generalization and specify that a section of the population that was not pretested be pretested. This will treated as an unintentional independent variable.
Interaction
The interaction between the selected variables and their treatment can affect the external validity. For instance, if the subjects are not selected from the population in a random manner, their characteristics are likely to jeopardize their own performance as well as the results of the study. Such results will not be fit to be used on the population or on another sample of that population.
Maximization of External Validity
Generally, the threats can be reduced if adequate measures are taken to ensure that the setting, sample and context are a true representation of the whole population. To attain this, the researcher must ensure that variables are well specified, experimenter effects well analyzed and selection interaction well managed. Additionally, effects of the experiment should be noted and arranged as well as the multiple treatment interference considered.
Conclusion
The internal and external validity are very important measures of the credibility and acceptability of any given research or study. Therefore, researchers should always seek to maximize the internal and external validity of their researches to ensure that they are acceptable. This is achieved by minimizing and if possible doing away with all the threats to both the internal and external validity.
References
Berg, K. E., & Latin, R. W. (2004). Essentials of Research Methods in Health, Physical Education, Exercise Science, and Recreation (2, illustrated, reprint ed.). New York: Lippincott Williams & Wilkins.
Chasteen, C. S. (2009). Internal and External Validity: Two Faces of the Same Coin. Oklahoma: University of Oklahoma.
Gravetter, F. J., & Forzano, L.-A. B. (2011). Research Methods for the Behavioral Sciences. London: Cengage Learning.
Mitchell, M. L., & Jolley, J. M. (2009). Research Design Explained (7 ed.). London: Cengage Learning.
