多元(复)回归分析（Multivariate regression analysis (complex)）.docVIP

多元(复)回归分析（Multivariate regression analysis (complex)） Multivariate regression analysis (complex) (Multiple Regression Analysis) The use of OLS (ordinary least squares) to do multiple regression analysis method is the most commonly used statistical research in sociology. The basic conditions of this method is a variable strain for a fraction of the type (variable interval scale measurement), and since the measurement variables scale without special restrictions. When the independent variable is the variable category, we can construct the number of categories (k) k-1 values of 0 and 1 dummy variable (dummy variable) to represent different categories. Therefore, if the proper use of words, multiple regression analysis is a very powerful tool. Multivariate regression analysis there are three main steps: First, using univariate and bivariate analyses to examine the various preparations into the regression analysis of the variables is consistent with the basic assumption of OLS linear regression analysis. - selected regression model, and evaluate the parameter estimation and goodness of fit (goodness of fit). In return we seriously consider the analytical results obtained before, should do the residual value (residuals) of diagnostic analysis (diagnosis). But we are usually set to determine the regression model (specification) is appropriate, will do in-depth analysis of the residual value. The first step is to examine each regression analysis will be included in the regression analysis of variable model. First of all, we must first determine whether the strain has enough variation (variability), and is close to the normal distribution (estimation of regression coefficient does not require strain is normal distribution, but the estimate of this assumption is required when determining the residual value for normal distribution. The strain state far left normal distribution when the residual value increases the possibility is not normal distribution). Secondly, the independ