SPSS data Analysis-optimal scale regression

In the linear regression model presented earlier, there is an implicit hypothesis that the arguments are continuous variables, but in fact, the independent variable is sometimes a categorical variable, similar to the factor in the analysis of variance, the categorical argument in the regression analysis, also by default as a continuous variable use, this will produce a problem, if it is unordered categorical variables, Then there is no difference between the different levels, each change of a unit, the impact of the dependent variable is the same, unable to analyze the trend, although the dummy variable can be used, but when the classification variable too much or the category of each variable is too high, this method is very cumbersome, in addition, when the category is more, There may be several categories that have a similar effect on the dependent variable, which is an analytical point, but the traditional linear model ignores this information, resulting in a waste of information. If it is an ordered categorical variable, then the variable encoding represents the high and low order of the variable, which has different effects on the dependent variable, while the traditional linear model still ignores this information and may lead to erroneous analysis conclusions.

Based on the above problems, statisticians have studied the optimal scale transformation method, which is specifically used to solve the problem of how to quantify categorical variables when modeling. The basic idea is based on the model framework that wants to fit, in order to ensure the relationship between the respective variables is linear, through a certain method of iterative iteration, to find an optimal quantitative score for the original classification variables, use this score instead of the original variable for subsequent analysis, so, not only the regression analysis, Any analysis method containing categorical arguments will be applied to this, greatly extending the scope of application of the analytical method.

The optimal scale transformation is used in regression analysis, which is the optimal scale regression, and the concrete process is

Analysis-regression-optimal scale

In this case, we want to analyze the influence of age, place of residence, education level on the number of children, the age is continuous variable, the residence is two categorical variable, the education level is ordered categorical variable, from the data situation, the type of the independent variable, age and place of residence can be directly included in the model analysis, The degree of education can be set in dummy variable form introduced into the model, but in this way, the equivalent of the variable is dispersed, not as a complete variable analysis, where we use the optimal scale regression.




All of these results are the results of regression analysis after the optimal scale transformation of variables, so how to see the transformation of variables? The process also provides a conversion diagram for reference, click the Draw button settings can be

SPSS data Analysis-optimal scale regression