Trinculo Files

#############################
# Multinomial speed testing #
#############################

## Overview

The speedtest.sh script generates some simulated multinomial genotype and phenotype data, and then analyses it using three implementations of a multinomial association test. The analysis is a standard likelihood ratio test testing if the variant has an effect on any of the phenotypes. 

You run the script as:

> speedtest.sh Npheno Nvar Nsample

Where Npheno is the number of phenotypes (not including the control category), Nvar is the number of variants to simulate, and Nsample is the total sample size (split evently across all phenotypes).

You will need to set the location of the sims and trinculo binaries in the script itself. 

The output, as generated by Trinculo, python and R is in trinculo.assoc.multinom, MNlogit_output.txt and nnet_output.txt respectively. Times for running each implementation are written to stdout.

## python
The speedtest.sh script calls the script testStatsmodels.py, which fits the model using the MNlogit function in the statsmodels models. You will need to have the statsmodels and pysnptools modules installed.

## R
The speedtest.sh script calls the script testNNET.R as an R BATCH command, which fits the model using the multinom function in the NNET package. You will need to have the nnet package (available from CRAN) and the snpStats package (available from Bioconductor) installed.