The Application Of The Clustering Analysis And Maximum Likelihood Analysis In The Premarital Sexual Behavior Study Of The Medical College Students

Objective: Premarital sexual behavior can not only affect the school work of the undergraduates but also result in premarital pregnancy and abortion. And it is associated with the increasing incidence of venereal diseases. Thus it induces a series of adverse effects such as seriously influencing the physical and mental health of the youth, leading great injury to both sides of the doers and bringing about heaven burden to the society. The routine investigation methods, which obtain the information from ordinary questionnaires can hardly get reliable and accurate data of the premarital sexual behavior [1].Using the randomized response technique to investigate the incidence of the premarital sexual behavior in college students; to classify the groups though clustering analysis method; to explore the related factors that may affect the premarital sexual behavior. Based on the above analysis, adopting the method of maximum likelihood to build a suit of diagnostic scale to discriminate the college students into different types in the future so that to provide the scientific foundation for assuring the generative health of the youth and the strategy-making of the related sections. Methods: 1 Sampling method Using the cluster sampling method: regard the 14 classes as 14 clusters (about 190 students in each cluster), and then randomly sample 3 classes from the clusters to carry out the preliminary investigation. 2 The sample size 2.1 Calculate the required number of clusters (k) according to the formula of calculating the size of cluster sampling introduced by Huang Ti Qian[2]. 2.2 On the base of the Q-type clustering analysis, to calculate the required sample size when comparing between different clusters. 2.3 Calculate the sample size according to the requirement of the discriminatory analysis. 2.4 Combine the three calculated sample sizes to determine the final sample size (n) and the number of clusters (k). 3 Technique route 3.1 Using "the questionnaires on the general conditions of the college students"to find out the factors that may be related to the premarital sexual behavior of the college students. 3.2 Adopting the three unrelated question randomized response model to investigate the incidence of the premarital sexual behavior in college students. 4 Statistical methods 4.1 Using the probability theory to analysize the data obtainedthough the randomized response technique and figure out the incidence of premarital sexual behavior in college students. 4.2 Evaluate the validity and reliability of the questionnaires on the general conditions of the college students[3]. 4.3 As for the information got from each suit of questionnaires, we use the Q-type systemic clustering method to get the final clusters of the samples. 4.4 Compare different variables among all the clusters, in which compare the sexual ratio and political features with the R×C chi-square test, when compare between every two clusters correct the significance level[4]; All the others variables are compared by the method of Kruskal Wallis test, Nemenyi method is used when compare variables between every two clusters, the calculate the data by the statistic software CS2000(Concise Statistics 2000). 4.5 On the basis of the clustering analysis, using the method of maximum likelihood, the conditional probability of every index and the correponding scores were worked out, thus establish the diagnostic Scale. Results: 1 Sample size 1.1 According to the formula of clusters sampling from limited clusters, the number of clusters (k) is 3. 1.2 From the view of comparing the characteristics among all the clusters, the sample size is 277. 1.3 According to the maximum likelihood method, the samplesize is 320. 1.4 Combines all the above results, the final size is 3 clusters and about 570 students. 2 The evaluation of validity and reliability Cronbach's αcoeffitient of the questionnaires was 0.8431, which indicates better Reliability of the study. Construct validity: through factoring analysis the factor loading was above 0.4, indicating that the frame of the scale...