When analyzing data using linear regression using the Ordinary Least Square (OLS) method, it takes an understanding of the assumption test that must be passed. The non-multicollinearity test is necessary to get the best linear unbiased estimator. The multiple linear regression OLS method has been widely applied in various fields: economics, agribusiness, and socio-economic fields. The estimation of the output of this linear regression has many benefits. Various research problems can be solved with this analytical approach. When we choose to use regression analysis, we are trying to see the influence or impact of one or more variables on other variables. Therefore, many researchers, lecturers, students, and practitioners choose linear regression using the OLS method as a data analysis tool.

Linear regression using the OLS method must meet the assumptions to obtain the Best Linear Unbiased Estimator (BLUE). One of the several assumptions required is no strong correlation or near-perfect correlation between the independent variables. This assumption is often called the non-multicollinearity assumption. Thus, if the independent variables have a strong correlation, the regression model that is built has a multicollinearity problem.

When we build a linear regression model so that the estimate is not biased, the assumption of non-multicollinearity must be met. Therefore, when we use linear regression with the OLS method, we must prove that the required assumptions are met. How to prove it? Yes, we can use statistical tests.

What test is appropriate when we prove that our model passes the multicollinearity test? As I said earlier, we need to return to our objective of conducting a non-multicollinearity test. We want to detect whether or not there is a strong correlation of the independent variables used in the model.

We can see the correlation coefficient between the independent variables. If you see a high coefficient value close to 1, it means that the independent variables have serious multicollinearity problems. However, apart from looking at the magnitude of the correlation coefficient between the independent variables, there is a test that many researchers choose, namely by looking at the value of the Variance Inflation Factor (VIF). We can get this VIF value easily using data processing software. The steps for the multicollinearity test and how to interpret them can be seen in the following video tutorial (video in Indonesian, please use an English translation):

If you have watched the video, at the end of the video, we discuss how to interpret it, especially regarding the decision-making of the VIF value. So, if the VIF value is small (less than 10), it means that the independent variable in the regression model passes the multicollinearity test. On the other hand, if the VIF value is large (more than 10), it indicates that the independent variable does not pass the multicollinearity test.

Well, there are two important points that we can take this time: (a) If the VIF is small and the correlation between the independent variables is low, it indicates that the regression model is free from multicollinearity problems (passing the non-multicollinearity test); (b) The higher the value of VIF and the closer to 1 the value of the correlation coefficient between the independent variables indicates that there is an increasingly serious multicollinearity problem in our regression model.

You may be wondering for those of you who have a high level of curiosity. Where did the VIF numbers obtain? If you want to know and you are more comfortable learning to use audio-visual, please watch the following video (video in Indonesian, please use the translation in English):

Suppose you have watched two videos at once. In that case, it means that you have at least a good understanding of the non-multicollinearity assumption test in linear regression using the OLS method. Why? Because you already know how to test non-multicollinearity either using data processing software or doing calculations manually. Please try it yourself using the data you have.

Hopefully, this article’s explanation can help you understand more deeply the terms of non-multicollinearity of linear regression using the OLS method. Hopefully, this article has a benefit for all of you. See you!