Timing simulations for BufferedMatrixMethods 1.3.5 (Simulation 2)

Written by Ben Bolstad
email bmb@bmbolstad.com

Background

A previous simulation found some small differences between the output of BufferedMatrix.justRMA() and RMAExpressConsole. It was theorized that the difference observed was due to small differences in the background correction implementations. The simulation here processes the same set of arrays (500) with the background correction turned off.

RunTest.sh Script for running RMAExpressConsole and R
R_Timing.sh Script for running and timing BufferedMatrix.justRMA()
output_500.settings Settings file for RMAExpressConsole (using 500 arrays as example)
Process500.R R script that used BufferedMatrix.justRMA() (using 500 arrays as example)
compareoutput.R Look at all the output and compare it.

Results

RMAExpressConsole_log.txt

R_log_500.txt

Discussion

The differences between the two implemtations are now at the order of 5*10^-7 as we would expect it to be given that RMAExpressConsole writes out expression values to 8 significant figures and using R we write to 16 decimal places. This shows that harmonizing the background correction code between the two implementations would remove any possible difference in the two implementations.