Regression analysis is a commonly used method in data processing, which can identify unknown relationships between data variables, obtain mathematical expressions that are more in line with variable relationships, and help users complete data analysis.

Next, we will introduce the multiple regression analysis method in regression analysis. IBM SPSS Statistics provides users with a mature multiple logistic regression analysis algorithm.

I. Overview

1. Data

Figure 1: Data Sample

This is a survey data on breakfast choices among different populations. Through multiple regression analysis using SPSS, population characteristic variables can be analyzed for breakfast types to find their relationships.

2. Functional location

Figure 2: Functional Location

Under the "Analysis" menu, we can find the "Multivariate Logistic" analysis in "Regression" and enter the window for multivariate regression analysis.

 

2、 Analysis method

1. Dependent variable setting

Figure 3: Dependent variable setting

The dependent variable is the quantity that follows the change of the independent variable, in this case referring to the variable of "preferred breakfast".

Click on 'Reference Category' and set the reference category for the dependent variable. This is the reference sample for analysis, and we set all categories to be compared with the last category in ascending order.

 

2. Factors and covariates

Figure 4: Factors and covariates

Factors can be simply understood as independent variables. Here, we will treat age segmentation, marital status, and lifestyle as dependent variables.

Covariance is a control variable that needs to be controlled during the analysis process and has a certain impact on the dependent variable. Here, it is set as gender.

 

3. Analyze the model

Figure 5: Analysis Model

There are three types of models available for multiple regression analysis in SPSS. The main effect refers to the analysis of the relationship between pre-set factors and covariates and the dependent variable; The full factor model includes both the main effect and the interaction analysis between factors and covariates; Customized step-by-step analysis allows users to define their own analysis types.

We can choose the main effect for analysis here.

 

4. Statistical settings

Figure 6: Statistical Settings

The settings in this window are for the statistical data analysis that needs to be performed, including multiple types of statistical data that can be selected. We have checked the pseudo R-square, cell likelihood, step summary, classification table, model fit information and fit, estimation under parameters (with a confidence interval set to 95%), and likelihood ratio test under the model.

Define subpopulation selection as a covariate pattern defined by factors and covariates.

 

5. Convergence conditions

Figure 7: Convergence Settings

In the condition dialog box, make convergence settings.

The maximum iteration number is the number of iterations that can be performed during regression analysis of data. This value must be an integer greater than or less than 100. The maximum step split is set as an equal fraction during iteration, and the system defaults to 5.

Logarithmic likelihood convergence can set a convergence value, and the logarithmic likelihood ratio function is greater than the set value during the regression process; The numerical settings for parameter convergence are similar.

In this example, the dialog box should remain default.

 

6. Option settings

Figure 8: Option Settings

Set the discrete metric to 'none' in the options dialog box.

The probability of data entering is 0.05 and the probability of data exiting is 0.1. Among these two parameters, the larger the former, the more data will enter the model; The smaller the latter, the more data is excluded, and both the entry and exit methods choose likelihood.

Leave the rest as default.

 

7. Save Settings

Set the variables to be saved in this dialog box, and if you need to output model information to an XML file, you can also do so in the following settings.

 

8. Complete the analysis

Figure 10: Result output

After completing the above settings, you can view the analysis results in the log output window