Introduction
According to a definition given in Business Dictionary, quantitative research implies “the use of sampling techniques (such as consumer surveys) whose findings may be expressed numerically, and are amenable to mathematical manipulation enabling the researcher to estimate future events or quantities” (“Quantitative Research”, 2015). It is one of the most fundamental and widely used ways of producing information in both social and mathematical sciences.
Threats to validity in quantitative research
However, the methods used to conduct quantitative research, and the data gathered in it may be invalid or faulty. According to Lærd Dissertation Guide, threats to the external validity of research are “any factors within a study that reduce the generalisability (or generality) of the results” (Lærd Dissertation Guide, 2012, p.3), and threats to internal validity “are things that make it difficult to prove that the independent variable is causing the changes in the dependent variable” (Boyd, n.d. par. 10). There are a few common threats to internal and external validity in quantitative research that are worth mentioning. For internal validity, they include experimenter effects and subject effects, compensation, compensatory rivalry, demoralization, diffusion (or imitation) of treatments, causal time order, experimental mortality, statistical regression, selection biases, instrumentation, testing effects, history effects and maturation. For external validity, these are history effects and maturation; the “real world” versus the “experimental world”; constructs, methods and confounding; selection biases. All of them should be taken into account when research is conducted so that the results would not be compromised.
Case study
A case study for this paper shows that the validity of the results can only be assessed under one condition – there is enough data. The problem statement is as follows:
In a published article about exercise, self-esteem, and happiness, the analysis shows, for women but not men, a high, positive correlation between the number of hours exercised per week and being childless or having no children under the age of 18 … the researcher concludes that women who have grown or no children are happier than those who have children at home. This is all the information given in a newspaper.
Although there is not enough information in the newspaper, it can be assumed that the research can be compromised by selection bias. However, this assumption can be proven faulty with the same statement that there is not enough information. According to a definition in a Research Methods Knowledge Base: “conclusion validity is the degree to which conclusions we reach about relationships in our data are reasonable” (Trochim, 2006, par. 3). It means that a conclusion can never be drawn if a researcher or interpreter has no information to rely on.
Response rate, missing data and errors they create
Other problems that may influence research results are the response rate and the missing data. When the first one is too low, it may threaten the results with a non-response bias, “the error resulting from distinct differences between the people who responded to a survey versus the people who did not respond” (How Response Rate Affects A Survey, 2015, par. 1). Firstly, the human factor impact should be excluded from the response collection. It may also prove useful to improve the speed or confidentiality of the research. There are a few techniques that allow dealing with missing data, too, for example, imputation or partial imputation, partial deletion, full analysis, and interpolation. There is another recommendation by Jason W. Osborne, the author of the book “Best Practices in Data Cleaning”:
Aside from examining missingness as an outcome … modern computing affords us the opportunity to fill in many of the gaps with high-quality data. This is not merely “making up data” … Rather, as my examples show, the act of estimating values and retaining cases in your analyses most often leads to more replicable findings as they are generally closer to the actual population values than analyses that discard those with missing data (or worse, substitutes means for the missing values) (2013, p. 130).
References
Boyd, N. (n.d.). What Is Internal Validity in Research? Definition & Examples. Web.
How Response Rate Affects A Survey. (2015). Web.
Laerd Dissertation Guide. (2012). Web.
Osborne, J. W. (2013). Best Practices in Data Cleaning. A Complete Guide to Everything You Need to Do Before and After Collecting Your Data. Thousand Oaks, CA: SAGE Publications, Inc.
Quantitative Research. (2015). Business Dictionary. Web.
Trochim, W. M. K. (2006). Research Methods Knowledge Base. Conclusion Validity. Web.