Mini-projects in Statistics

Index of This Page

  		• Projects in Time Series Analysis;
  		• Projects in Categorical Analysis;
  		• Multi-Linear Regression Analysis;
  		• Bayesian Analysis.
  		• Generalized Linear Regression/Anova Analysis.

    Time Series Projects (S and S plus is mainly used):

     
  		•  click here to see some source codes;
     
  		•  click here to see the analysis of the output.
    Back to Top

    Categorical Data Analysis Projects (SAS is mainly used):

     
  		• Create tables for categorical data analysis (Use of Weight, Freq, table, Relative risk);
     
  		• More option of creating tables for categorical data analysis(Use of table, freq, Measures to calculate Gamma, Lambda Asymmetric, Uncertainty coefficient);
     
  		• Categorical Data Analysis using of table, freq to calculate Pearson chi-squared statistics, likelihood ratio --Independence test;
     
  		• Categorical Data Analysis using of LOGISTIC Model (use of Logistic: explanatory variables are quantitative);
     
  		• Categorical Data Analysis using of chi-squared tests for independence, Calculating Cochran-Armitage linear trend test (More options of "table": ChiSq, trend);
     
  		• Categorical Data Analysis use of logit model (use of catmod, weight, direct: explanatory variables are qualitative and quantitative);
     
  		• Loglinear Model (Use of Table, Freq, Loglin, Catmod, plot);
    Back to Top

    Multiple Linear Regression Analysis(SAS is mainly used):

     
  		• Project Report ;
     
  		• Click here to see the data (Price of beef vs. various factors;
     
  		• Click here to see the code1.sas(Find some candidate models use selection criteria, backward and stepwise selection procedure, and analyze the collinearity between the regressors);
     
  		• Click here to see code3.sas(Compare several candidate models to see which one is the best);
     
  		• Click here to see code4.sas(Residual Analysis);
    Back to Top

    Bayesian Analysis (C/C++ was used):

     Brief Descritpion:In this project, I used Monta Carlo Markov Chain to generate posterior estimator,
    to prove that if the joint posterior distribution is improper, the Gibbs sampler
    still provide some reasonable result instead of giving us some alert. So we must
    know the properiety of the jointly posterior distribution before we use Gibbs Sampler.
     
  		• Project Report (Click here to download a .ps file);
    Click here to download Ghostview to open the .ps file
     
  		• Algorithm1(Here are the programs to generate the posterior histogram based on the frequestist random effect one-way anova model using improper priorsc/c++);
     
  		• Algorithm2(Here are the programs to generate the posterior histogram based on the Hiearchical Centering one-way anova model using improper priors;c/c++);
     
  		• Algorithm3(Here are the programs to generate the posterior histogram based on the frequestist random effect one-way anova model using proper priorsc/c++);
     
  		• Algorithm4(Here are the programs to generate the posterior histogram based on the Hiearchical Centering one-way anova model using proper priors; c/c++);
     
  		• Generate random values from gamma(alpha, beta)(c/c++);
     
  		• Generate random values from N(0,1)(c/c++);
     
  		• Generate random values from Uniform(0,1)(c/c++);
     
  		• Generate some artificial data(Generate some data for our analysis);
     
  		• File to create plots of posterior distributions(Histograms and smoothed histograms);
    Back to Top