R programming for bayesian analyses

R programming for bayesian analyses


Job Description

Immediate need for experienced R and BUGS (OpenBUGS preferred) programmer to develop code providing our team with ability to perform statistical analyses of data sets (using a standard input format).

Focus on meta-analysis (using R Metafor package for example) and application of Bayesian modeling (BRugs or equivalent to call OpenBUGS), in order to allow us to more efficiently conduct multiple analyses (rather than developing custom code each time).

This is a "work for hire" arrangement, with ExVivos owning the work product. Only independent consultants need apply. For qualified proposals, we can discuss the product specs provided after putting in place non-disclosure agreement before committing to the payment/project.

EXVIVOS will consider bids on project basis more favorably but is willing to evaluate hourly bids as well. Time line is for completion of coding by end of July 2013.

More Specifics:
These tasks are listed in a manner to emphasize the desire for writing the code as separate functions that work together and provide flexibility to enhance capabilities with additional functions later (such as adding ability to perform analyses on "patient-level" datasets). Software must meet these general requirements:

General description of work product:
1. Work product is computer software written in the R programming language, with control of OpenBUGS software for MCMC Bayesian analyses, along with instructions for ExVivos to operate independently, suitable to execute through R-Studio interface within Windows 7 environment (running in Parallels v8 VM on Mac).
2. Work product will include sufficient documentation within it to enable ExVivos to identify and use the functions contained within the work product.
3. Contractor will identify and provide for EXVIVOS R code for the priors that are most appropriate for the biologic and clinical data of interest. [For Contractors less expert with statistics, we can be flexible here and provide this information from alternative sources.]
4. EXVIVOS will provide standard input file to Contractor (to use csv file format similar to Cochrane Collaboration).

Additional specific functional requirements:
1. Select file for analysis
2. Manage variable names - based on using the input format provided as the csv file from Cochrane
a. remove spaces
b. contract long names
c. add variables that will be necessary, such as ones to place the yi and vi we calculate so these can be compared to those provided when they are provided
3. Select cases
a. This could be based on the group numbers provided, but likely will need to be case by case selection
4. Select priors etc, as defined in the following section, though perhaps best to create separate and callable function
5. Select analyses to be performed
a. In a Cochrane based analysis, will want to verify results of the meta-analysis performed relative to those provided by Cochrane (limits of agreement, correlations?)
b. meta-analysis
c. Bayesian modeling
- selection of priors
i. dichotomous and/or continuous
ii. skeptical, non-informative, optimistic and/or user-defined
iii. setting parameters (default values sensible?)
d. Bayesian prediction
i. dichotomous and/or continuous
ii. skeptical, non-informative and/or optimistic
iii. setting parameters (default values sensible?)
e. Meta-regression
i. variable selection
ii. weighting
6. Plotting results for each analysis

The final work product will require validation using up to four datasets selected by EXVIVOS.