Comparing Study About Statistical Methods Which Were Applied To Analyze The Relationship Between Social-Psychological Factors And The Internet Use In College Students

Objective This paper used reliable measuring scales, reasonable variables and statisticalmethods to investigate the social and psychological factors on the internet addiction ofcollege students. Find a suitable technique to solve the problem and hope it wouldprovide some useful information on the methodology in the similar research.Methods The college students pursuing bachelor's degree enrolled Fuzhou University andFujian Normal University were investigated by means of proportional stratified clustersampling. The data was analyzed by Analysis of Covariance, Multiple StepwiseRegression Analysis(MSRA), Principal Component Regression Analysis(PCRA), RidgeRegression Analysis, Stepwise Discriminate Analysis, Partial Least Squares Analysis,Multilevel Models and so on.Results The population of this study is the college students of two comprehensiveuniversities (Fuzhou University and Fujian Normal University). The students are askedto finish the questionnaires in class. The number of the total utility questionnaires is 645,and there are 281 male(43.6%)and 364 female(56.4%).Use Partial Least Squares Regression to analyze the data and Predicted Residual SSto compare the prediction effect, the result showed: the equation isY=40.78+4.068gender+5.98condition+0.286pressure-0.305support+0.141depress+0.066loneliness-0.25esteem. Partial Least Squares Regression is the best multiple regressionmethod to predict the model. Use HLM5.04 to carry out Multilevel Models analysis. Theresult showed: there is aggregation in class level. In lower level, the factors impact theinternet use were gender, pressure, social support, depress, loneliness and esteem. Inhigher level, the factor impact the internet use was the condition of browse on internet.Conclusions The internal and external factors impact the internet use in college students aregancer, the condition of browse on internet, pressure, social support, depress, lonelinessand esteem. The proportion rate of internet addiction can be prevented and controlled byinterfering in these aspects. It's acceptability to solve multicollinearity by some methods, such as PCRA andRidge Regression Analysis. But there are some flaws. In some conditions, other moreeffective methods, such as Partial Least Squares and Multilevel Models can be used tosolve the problem of the data structure and multicollinearity.