Dummy Variables in Regression Analysis

In regression analysis, a dummy variable is a numerical value that is artificially created and assigned to a subgroup of an independent variable that can be represented in two or more ways that a research wishes to determine how they affect a dependent variable. The concept is critical to regression analysis because how this type of statistical analysis works. Since in regression analysis all independent variables are represented in numerical values and treated as ratios of each other, it is possible to meet a challenge where variables are assigned numerical values but the numbers do not denote anything about the variable ( Skrivanek, 2009). For example, in a regression statistical analysis on the relationship between football match watched in a weekend and the gender of the attendants, the “football matches watched during a weekend” can be represented by the dependent variable Y and “gender” by the independent variable X. In such a scenario, two dummy variables can be created with either one having a value of 0 or 1. The resultant dummy variables can be represented as follows; D= 1 if male and D = 0 if female. 

Dummy variables are important to regression analysis. Notably, they enable the use of a single equation in the analysis of more that two subgroups by eliminating the need for multiple equations for each subgroup. This is of critical importance because in regression analysis because in most real-life cases, the independent variables can be represented in two ways hence the use of artificially created values to represent the ways in which the variable can represented. For instance, in the example given above, instead of creating two equations of each time for the male and female subgroups, the dummy variables D=1 if male and D = 0 if female enables the researcher to have only one equation and have male attendants represented by 1 and female attendants represented by 0. In regression analysis, the coefficient of the independent variables has a maximum of 1. However, in special cases where the variables have a multicollinearity, then the coefficient can rise to above 1. Therefore, in the example given above, if coefficient of the independent variable is found to be 1.6 after running the regression, that would mean that there is a high correlation between the independent variables. 

Reference 

Skrivanek, S. (2009). The use of dummy variables in regression analysis.  More Steam, LLC .