README file for parallelized Gaussian Mean Change Point Detection

Contact (unless otherwise noted): nsong@alumni.princeton.edu (C) Nancy Song, Princeton University

Overview: this program executes a parallel version of change point detection of changes in the Gaussian mean. Note that for fluorescence spectroscopy, input are the intensities at each time point, not photon counts.

Requirements: GNU C Library 2.12.2+, OpenMPI 1.4.2+

Instructions: (1) Build Parallel-GM program using the makefile (2) Run Parallel-GM using the following arguments: filename, alpha, beta (a) alpha is the Type-I error (b) beta is the confidence region. Only accepts values from {0.69, 0.90, 0.95, 0.99} (3) Output is a file with the name of the file appended with .cp that has the time index of the change point detected and the left and right confidence intervals

Example: The following runs Gaussian mean change point detection on the file file1 with two processes and a Type-I error of 0.05 and a confidence interval of 0.95

    Type into command terminal:
    mpirun -np 2 Parallel-GM filename 0.05 0.95

    Output file would be file1.cp
    Sample from file1.cp:   
    99 97 104

    This indicates that a change point was found at the 99th time step. 
    Confidence region is from index 97 to 104