Dr. Debajyothi Sinha
Ron & Carolyn Hobbs Endowed Chair/ Professor, Department of Statistics.
Research interests
- Survival analysis and Bayesian analysis.
- Bayesian Biostatistics.
- Modeling Cancer prevention data.
- Cure rate and survival data.
- Modeling Cancer relapse data and recurrence data.
- Semiparametric empirical Bayes.
- Methods for skewed and heteroscedastic response.
- Equivalence Testing
- IMR Prior
- Longitudinal Analysis
- Selection Methods
- Univariate Analysis
- NIH papers
Main Code for computing the empirical power and type I error for Proportional Odds Model when the data are generated from proportional odds model or proportional hazard model respectively.
Effect Size-II One of functions we used in calculate the MLE of effect size beta.
Newton-Raphson Effect Size A function for using Newton-Raphson algorithm to estimate maximum likelihood estimation (MLE) of beta of under proportional odds model. The input include patient survival time, censored status and covariates. The output is the point estimate for the covariate effects beta.
Effect Size-I One of functions we used in calculate the MLE of effect size beta.
Univariate Proportional Hazard A function for generating data from univariate proportional hazard model. The input includes the number of samples n and the true effect size beta. Then output data are the survival times for every patients..
Univariate Proportional Odds A function for generating data from univariate proportional odds survival model. The input includes the number of samples n and the true effect size beta. Then output data are the survival times for every patients.
COV effects estimator A function for calculating the standard deviation for the estimation of covariate effects beta.
Cure rate-IMR R code used for IMR prior for Cure Rate survival model. Given the survival data, the output for the function includes the posterior samples for the covariates effects using IM prior given the input data.
PH-IMR R code used for IMR prior for proportional hazard model. The baseline hazard function is assumed to be piecewise constant function. Given the survival data, the output for the function includes the posterior samples for the covariates effects using IM prior given the input data.
G-prior R code used for gprior for proportional hazard model. Given the survival data, the output for the function includes the posterior samples for the covariates effects using IM prior given the input data. This method is used as a reference method to evaluate the performance of IMR prior.
Normal prior R code used for Gaussian prior for proportional hazard model. Given the survival data, the output for the function includes the posterior samples for the covariates effects using IM prior given the input data. This method is used as a reference method to evaluate the performance of IMR prior.
PH generator Generates data from proportional hazard model with piecewised baseline hazard function. We used these data in our simulation study. The input includes the covariates matrix, the coefficients for covariates, number of sample size, and the baseline hazard function. The output is the survival time for all patients.
The code below corresponds to methodology discussed in the following paper:
Bayesian Partial Linear Model for skewed longitudinal Data
[Status: Submitted to Journal of the American statisticial assosciation on 12/12/12]
The zip folder contains scripts (JAGS,R) for the following:
Simulation Study:
R code is for generating data, theJAGS code is for the model and prior.
Data Example:
R code is to read the data, reshape the data, standardize the data. JAGS code covers the model and prior.
Longitudinal
The code below corresponds to methodology discussed in the following paper:
Bayesian variable selection for skewed heteroscedastic error
[Status: In progress]
The zip folder contains scripts (JAGS,R) for the following:
Simulation Study:
R code for generating data, JAGS code for model and prior.
Data Example:
R code for cleaning, reading data, JAGS code for MCMC update.
Variable selection
The code below corresponds to methodology discussed in the following paper:
Bayesian regression for univariate skewed heteroscedastic error
[Status: In progress]
The zip folder contains scripts (MATLAB) for simulation studies and a data example.
Univariate
The code below corresponds to methodology discussed in the following paper:
Empirical Bayes estimation for additive hazards regression models.
Sinha D, McHenry MB, Lipsitz SR, Ghosh M.
[Status: Published; Biometrika. 2009 Sep;96(3):545-558. Epub 2009 Jun 26.]
Empirical Bayes Full Additive Bayes