What are the different types of hypotheses? Why might a researcher use more than one type?
Establishing a well-grounded hypothesis is crucial for successful scientific research. It shows the researcher’s competence in his or her field of studies, indicates the presence of unique ideas and innovative suggestions and defines the overall direction of further research (Ary, Jacobs, Razavieh, & Sorensen, 2009). A hypothesis is called inductive if it is build based on the generalization of some phenomena and observed relationships. A deductive hypothesis is the one derived from a theory. Usually, three types of hypotheses are defined, namely research, null and alternate. The first one “states the relation one expects to find as a result of the research” (Ary et al., 2009, p. 91). The null hypothesis states that there exist no relationship between the variables or that the expected result will not be proved. The alternative hypothesis, on the contrary, represents the positive outcome of one’s suggestions, indicating that the relationship between the variables exists (Ary et al., 2009, p. 92).
Using more than one hypothesis can help a researcher to explain the phenomenon more comprehensively, to consider all possible options for the solution of the problem and to reach some meaningful result (Gliner & Morgan, 2009). A researcher, thus, can show more unbiased and objective approach to the problem.
What parametric and nonparametric tests are used for hypothesis testing?
Parametric tests require numerical data and researcher’s assumptions about the selected parameters; the data under consideration should be from ratio or interval scale (Gravetter & Wallnau, 2011). Parametric tests usually ensure more precise testing of alternative hypotheses, as they provide the comparison of null and alternative hypotheses based on the calculated numeric probability (Sheshkin, 2003). Among the key parametric tests are the t-test for one or two independent samples, single-factor between-subjects variance analysis and one or two sample proportion test (Sheshkin, 2003).
Nonparametric tests involve the group of statistical criteria, which do not include the probability distribution parameters in the calculation and are based on operating with frequencies or grades. Those tests deal with nominal and rank-order data (Gravetter & Wallnau, 2011). The most widely recognized nonparametric tests are Friedman analysis of variables test, Kolmogorov-Smirnov two-sample test, Wilcoxon signed-ranks test, Moses test for equal variability, Siegel-Tukey test for equal variability, chi-square tests for homogeneity and independence (Sheshkin, 2003).
What is the relationship between hypothesis testing and data?
Curran (2011) defines statistical hypothesis testing as a “formal, objective, repeatable way of making inferential statements based on data” (p. 98). The available data, thus, defines the methods of hypothesis testing. It may be impossible to state hypotheses using specific, measurable numeric parameters. Such situations are typical, for example, for humanities or political sciences (Gravetter & Wallnau, 2011). The available data can fall into several categories. First, the data can be of the nominal type with a variable having only name value.
It is important to establish the similarity or difference between the objects based on some qualitative criteria; there may exist only to variables 0 or 1. Second, there can be ordinal data, when the variable is measured based on rank or order. In the case of interval data, there can be some “value on a continuous scale but there is no zero point where the trait does not exist” (Black, 1999, p. 52). Finally, the ratio data type suggests that there can be a zero value on the continuous scale, which indicates the complete absence of some characteristics. For the nominal and ordinal data, nonparametric hypotheses testing is used while the interval and ratio ones can deal with parametric testing.
References
Ary, D., Jacobs, L., Razavieh, A., & Sorensen, C. (2009). Introduction to research in education. Belmont, CA: Cengage Learning.
Black, T. R. (1999). Doing quantitative research in the social sciences: An integrated approach to research design, measurement, and statistics. London, England: Sage Publications.
Curran, J. (2011). Introduction to data analysis with R for forensic scientists. Boca Raton, FL: CRC Press.
Gliner, J., & Morgan, G. (2009). Research methods in applied settings: An integrated approach to design and analysis. New York, NY: Routledge Academic.
Gravetter, F., & Wallnau, L. (2011). Statistics for the behavioral sciences. Belmont, CA: Cengage Learning.
Sheshkin, D. (2003). Handbook of parametric and nonparametric statistical procedures. Boca Raton, FL: CRC Press.