Skip to content.
CCB > ProgressReport > CCB_Renewal > CCB_Roadmap2007 > CCB_Roadmap2007_UnsolvedProblems > CCB_Roadmap2007_UnsolvedProblems_Stats01

CCB Roadmap (2007) Unsolved Problems

Implement value, vector and tensor uni/multivariate statistics for population studies

Overview

  • This project aims at developing stand-alone tools for statistical analysis of group differences using displacement (warping) fields, based on the mathematical models and Matlab implementations developed by Paul Thompson and Natasha Lepore.

Requirements

  • Univariate measures (Mostly exist - see Pipeline, T-tests and ANOVAs)
  • Vector-valued stat maps:
    • This tool should be invoked via AIR-like command-line syntax
      • stat_maps -nog number_of_groups NumSubjGrp1? ... NumSubjGrp? -grp1 Grp1Fld1.fld ... Grp1Fld_k1.fld [-grp2 Grp2Fld1.fld ... Grp2Fld_k2.fld] -output outputFileName [options]
    • Stat maps should include (Hotelling's T2, log(JTJ), and variaties of the maps reported in the pubs below)
    • The Pipeline and Grid binaries for the prototype version of the tool are available (via the Pipeline environment and CCB SVN (Natasha's Matlab code)).
    • A tested Fortran version of the multvariate Hotelling's T2 statistics is built and available on the Grid (cerebro-rcc.loni.ucla.edu) at: /cxfs/ccb/CCB_SW_Tools/others/Statistics/HotellingT2Statistics, and available via the Pipeline.
  • Extended stat maps to DTI tensors (talk to NathanHageman, PaulThompson, Ming-Chang Chiang)?

Status

  • Version 0.9 delivered and and available through the pipeline, integrated with KL_MI registration (full testing cycle required).

Pipeline Usage Documentation

References