Latent Class Analysis Of SNPs Distribution Characters In Depression Patients

Latent class model(LCM) forms a theoretical system in the development of categorical latent variables analysis. LCM integrated structure equation model with linear model, formed its own advantages, its purpose lies in the minimum number of latent clusters to explain the link between manifest variables. The introduction of LCM not only made up for the deficiencies of SEM that only deal with continuous latent variables, but also provided medical or social science researchers with a new thought. The advantages of LCM adapted the requirements of the statistical methods improvement because medical model changes cause the complexity of the etiology . Therefore, in medical research into the relationship between multivariable introducing LCM has the important practical significance.In this paper the principle and methods are systematically presented, especially the analysis procedures of LCM, including model parameterization, parameter estimation, model identification, goodness-of-fit evaluation, classification and interpretation of the results. Model parameterization including probabilistic parameterization and log-linear parameterization, and parameters of the model are explained; iterative maximum likelihood estimation methods, namely EM and NR algorithms, were used in parameter estimation; goodness-of-fit evaluation of the model should combine Pearson chi-squares, likelihood ratio statistics, AIC, BIC to assess; Finally, observations ought to be classfied into the appropriate latent clusters and explained correctly. Including exploratory and confirmatory two basic types, this article described the basic principles and the analysis steps, followed by the introduction of multi-sample latent class modeling(MS-LCM).Data simulation is carried out by the module of Monte Carlo simulation study with Mplus software. First of all, generate simulation data in Monte Carlo simulation study modules according to the designated model, then import directly it to Mplus programme to implement LCM analysis and multi-sample comparison, meanwhile show it clearly on the plot. Two kinds of LCM are simulated and analyzed,namely two-categorical and three-categorical LCM ,including multi-sample analysis. In example analysis, SNPs (single-nucleotide polymorphisms) data of depression disorders are analysized by LCM, and 7 groups of SNPs data are compressed and reduced to find out the latent distribution of SNPs. AIC and BIC implied that 7 groups of SNPs data are divided into two latent clusters, and the probability of each latent clusters is 22.378%, 77.622%. The first class tends to heterozygote of HTR1B gene;the second class tends to homozygous of HTR1B gene. Eventually, to explore whether the clinical symptoms or some psychological traits related to it so as to achieve the evaluation of gene function. The results showed that cluster-1 had a low-tendency to negative coping style, while the cluster 2 have a high-tendency to negative coping style.The model parameterizaton, parameter estimation, model identification, goodness-of-fit evaluation, observations classification and so on are explored, advice on latent class analysis is given discussion.