In the field of data analysis, the VIF test of SPSS can help researchers determine whether there is a high degree of collinearity between multiple variables, thereby screening and deleting data sets with strong collinearity, and optimizing and improving the regression equation model of relevant data.

1、 How to perform VIF test in SPSS

If the data we collect contains multiple independent variables and a dependent variable, and we want to understand whether there is a high degree of collinearity between these independent variables, we can use the SPSS regression equation method for VIF testing. It should be noted that the VIF test of SPSS has certain requirements for the data variables. The dependent variable is a continuous variable and the independent variable is two or more continuous variables. Only when the above conditions are met can the VIF test function run normally.


1. The following figure shows the R&D investment data for a certain mobile phone chip component. Given the instrument accuracy as the dependent variable, it is necessary to test whether there is a high degree of collinearity between the three data sets of R&D duration, average annual loss per unit, and annual R&D investment. So first, find the functional module of SPSS regression analysis, click on the 'Linear' button, and enter the instruction operation page of VIF test.

Figure 1: Development data of a certain mobile phone chip

2. On the application function page of linear regression, first put the data of instrument accuracy into the content box of the dependent variable, and then put the three sets of data of annual R&D investment, annual average loss per unit, and R&D duration into blocks to complete the variable option setting of collinearity test. This way, the relationship between the three sets of variables of annual R&D investment, annual average loss per unit, and R&D duration can be judged based on the subsequent statistical results.

Figure 2: Instrument precision as dependent variable

3. Next, enter the statistical module of linear regression, select the estimated value in the regression coefficient display column, and then check the four options of collinearity diagnosis, description, R-squared change, and model fitting on the right side of the figure. Select the Debrecen Watson mode in the residual content box, and then you can obtain the corresponding results such as model summary, single factor analysis, and collinearity diagnosis.

Figure 3: Selecting Estimated Values

4. In order to check whether the development data of the mobile phone chip conforms to a normal distribution as a whole, we need to select the DEPENDNT mode in the linear regression graph settings and choose a standardized residual graph presented in the form of a normal probability graph.

Figure 4: Normal distribution and residual plot

2、 How to view the results of SPSS VIF test

The results of SPSS collinearity diagnosis mainly include residual plots, model summary R-squared values and significance values, one-way analysis of variance tables, regression model coefficients, and other statistical results. If the VIF value is greater than 5, it indicates a high degree of correlation between various variables, and relevant data needs to be considered for exclusion or replacement.

1. In the regression standardized residual plot with instrument precision as the dependent variable, the numerical points are basically arranged on the straight line in the figure below, indicating that the R&D data of the mobile phone chip conforms to a normal distribution. The relationship between the three variables of annual R&D investment, average annual loss per unit, and R&D duration can be determined based on subsequent model summaries, analysis of variance, regression coefficients, and other results.

Figure 5: R&D data follows a normal distribution

2. Next, let's take a look at the SPSS model summary table. The R-squared value is 0.981, and the adjusted R-squared value is 0.978. The closer the R-squared value is to 1, the better the fit of the data model. The Debusson value is 0.871, with a significance value less than 0.001, indicating that at least one of the variables, such as R&D duration, annual R&D investment, and average annual loss per unit, can significantly predict changes in the dependent variable, instrument accuracy.

Figure 6: The R-squared value is 0.981

3. In the ANOVA table of mobile phone R&D data, the dependent variable is instrument accuracy, and the predictor variables are R&D duration, annual R&D investment, and average annual loss per unit. The sum of squares of the regression is 5420.998, and the mean square is 1806.999. A significance value less than 0.001 indicates that the data model construction is reasonable and can be further judged based on subsequent analysis results.

Figure 7: Regression sum of squares and significance values

4. In the final coefficient table, the standardized coefficient of annual R&D investment is 0.562, with a significance value of less than 0.001. The standardized coefficient of annual average loss per unit is 0.470, with a significance value of less than 0.001, indicating that the variables of annual R&D investment and annual average loss per unit have a significant impact on instrument accuracy. At the same time, the VIF values for annual R&D investment and average annual loss per unit were 5.476 and 5.845, respectively, indicating a highly correlated relationship between the two variables.

Figure 8: The VIF values for annual R&D investment and average annual loss per unit are too high

The above is the answer to how to perform VIF test in SPSS and how to interpret the results of SPSS VIF test. If it is necessary to understand whether there is a high correlation between multiple continuous numerical variables, SPSS regression analysis can be used to calculate VIF values for further judgment and analysis.