Flowcharts A, B, C, and D provide decision trees to direct the researcher in identifying the correct statistical methodology for the specific setting of their investigation. An explanation for how to choose between one and two or more factors, parametric and non-parametric methods, as well as one and two or more groups can be found below. Once a specific methodology has been identified, the reader can use the navigation bar on the left to access the description and examples for that statistical method.
Distinguishing between One-Factor and Two or More-Factor Analysis
The one-factor methods apply to settings where the researcher is studying the effect of one factor on a response variable. On the other hand, if the researcher is interested in studying the effects of two or more factors on the response variable, then methods from the two or more-factor analysis should be employed.
For example, a researcher is interested in studying the effects of a new teaching technique on the ability of dentistry students to correctly recognize tooth color. Students are randomly assigned to a control group that is taught using a classical method to recognize color, and to a treatment group that is taught using this new technique. After the training, students are assessed on their ability to recognize the color of given tooth specimens. If the researcher is interested in studying the effect of one factor—teaching technique, on the response variable—the ability of students to recognize color, then this is a one-factor setting. On the other hand, if the researcher is interested in questions on how two or more factors influence s student’s ability to recognize color, then a two or more-factor analysis should be employed. For example, the researcher might be interested in studying how the teaching method and gender affect color recognition, since the ability to recognize color is strongly associated with gender. This would fall under the two-factor analysis setting.
Deciding between Parametric and Non-Parametric Tests
When deciding to use a parametric versus a non-parametric test, the reader should take into consideration the specific assumptions underlying these methods. Parametric methods require normally distributed populations or a large size of the sample, and some methods require knowledge of the population standard deviation. The non-parametric methods do not have these requirements, and so can be applied to a broader variety of scenarios. However, the tradeoff is that the non-parametric methods are not as efficient as the parametric tests, and require stronger evidence to attain a significant result, such as rejecting a null hypothesis.
Distinguishing between One-Group and Two or More-Group Settings
The one-group methods apply to settings where the research question is about a single population. On the other hand, if the researcher is interested in comparing two populations, then methods from the two-group setting should be applied.
For example, suppose the goal of the researcher is to understand the effect of irradiation on the flexural strength of a composite dental material. If the question is to determine whether irradiation affects the strength of a particular composite material or estimate this effect, then a one-group method should be applied. If on the other hand, the researcher is interested in comparing the effects of irradiation on two different composite materials, then the two-group setting methods should be used. To compare 3 or more composite materials, the One-way ANOVA should be used. A two-group setting also arises when the goal of the researcher is to compare the effect of 20 days of exposure to irradiation to that of 40 days on the same composite material. Here Group 1 consists of the composite specimens exposed for 20 days and Group 2 of the specimens exposed for 40 days. If the measuring process does not destroy the specimens, then it is possible that the same specimens are measured at day 20 and at day 40. So even though they are the same physical specimens, they form different treatment groups. In this setting, a paired comparison test can be used.
Flowcharts
