Chapter 12 of Statistics for Criminology and Criminal Justice examines a hypothesis-testing technique called “analysis of variance” (ANOVA) that can be used when the independent variable (IV) is categorical with three or more classes and the dependent variable is continuous. It tests statistics called “F statistics” due to its F probability distribution. ANOVA combines variances and means into one specific test for significant differences between means.
The chapter provides five main arguments to use the analysis of variance technique. First, analysis of variance should be used in cases with category IV to eliminate any increase in Type I error rate that can emerge if a series of tests will be applied. ANOVA is not a viable analytic strategy due to familywise error (Jacinta, 2012, pp. 281). Second, the analysis of variance technique ensures an easier test than complicated a series of t-tests. For example, there is a category IV with classes A, B, C, and D. A researcher needs to run a separate t-test for each combination of classes that will be finalized with the results that are difficult to interpret (Jacinta, 2012, p. 282). Third, ANOVA is also a useful analytical technique due to its types: between-group and within-group. Another argument for the ANOVA technique in the chapter is that it does not indicate the size of the effect, so it applies the omega squared measure (Jacinta, 2012, p. 287). Lastly, the F statistics technique is effective to make conclusions regarding a null. For example, when after a test, if F distribution is concentrated on the left side of zero, obtained values of F will be positive (Jacinta, 2012, p. 284).
There are several limitations of the chapter associated with the ANOVA technique. First, the chapter does not consider that the technique is designed to test against an alternative hypothesis to the null, thereby maybe not be suitable against a specific hypothesis. Second, the chapter does not include cases with several factors and dependent variables.
Reference
Jacinta M. Gau. (2012). Statistics for Criminology and Criminal Justice. Sage.