2. A Few Keywords and Definitions for Understanding Statistics – VII (contd.)

2.10 Research Methods (Contd.)    

Too often it's not just enough to be able to describe an event or a phenomenon, but also to study the relationship between two different entities or properties or to predict one variable from another. Such predictability provides a more accurate understanding of phenomena and, at the same time the power to construct evidence-based theoretical premises and the ability to explain processes, behaviors, and phenomena better. These methods may be termed as predictive or relational research methods and there are broadly two methods that allow such power and predictability namely the 1) the Correlational methods, and 2) the Experimental methods.

The Correlational method assesses the extent of the relationship between two measured variables. If there is a considerable correlation between two variables, then it is possible to predict one variable from the other with a fair degree of accuracy. For example, there is a high correlation between atmospheric humidity and rain; height and weight; or cholesterol level and incidence of heart attacks. Based on the knowledge of the humidity, height, or cholesterol level, we can predict the amount of rainfall in a given region, the height of a person, and the probability of heart attack in a given population. However, correlation does not imply causation. Thus, we cannot say based on high correlation that one variable is caused by the other or that a change in one variable is a result of a change in another variable. We can’t say for instance that an increase in weight is caused by an increase in height, or that high cholesterol has caused a heart attack. This is a drawback of correlational methods. Correlation can be positive or negative and thus when there is a negative correlation, an increase in one variable may cause a decrease in the other. For example, when height or elevation from the ground increases, the temperature decreases, thus, there is a negative correlation between the two.

  To some extent, the lack of causation is compensated for in the correlational method by regression of a single variable on one or more predictor variables guided by theory or past knowledge. A regression method may be deemed as an extension of primarily the correlation method, whereby a variable is explained or predicted to a greater extent and more accurately in terms of other variables. However, the predictability enabled by regression also does not imply causation. The correlational methods are thus essentially predictive.

The Correlational methods are essentially observational in the sense that they do not interfere with the conditions surrounding a subject or set of subjects and their traits. Unlike the Correlation methods, however, the Experimental methods are designed to control and if required, manipulate the conditions surrounding the subjects and variables of interest. The idea is to mitigate undue influences and redundancies emanating from unwanted or irrelevant variables. The reason behind the attempt to mitigate the undue influences and mistaking irrelevant variables as being associated with the phenomena being studied is to single out the real cause of the phenomena. Thus, the experimental research methods may be deemed as explanatory in the real sense, whereby the real explanation or explanatory variables and the nature and extent of the causative and explanatory relationship are established.