restinginpca

Short descriptions about the objectives

• Maybe we can add descriptions about where to find things in this repository?

Analysis in this repo requires the some R packages not on CRAN

MexPosition did not update to be on the current CRAN, but can still be downloaded from Github. PTCA4CATA with parallelized bootstrap is on JY's github. superheat that enables reversing of axis when plotting suqare matrices are on MC's github

devtools::install_github("cran/MexPosition") devtools::install_github("juchiyu/PTCA4CATA") devtools::install_github("mychan24/superheat")

Notes:

• 2019.9.23: HMFA-normalization fixed.

  • This was only fixed for the 4rows analysis, not for the original MuSu.

  • The original version extracted both levels of singular values from the original data table. However, these two levels of singular values should be extracted hierarchically: extract delta for edges -> weight the edge tables -> extract delta for subject -> weight the subject tables

• 2019.9.25: Contribution fixed

  • This was only fixed for the HMFA-edgetype (7 and 11) in MuSu_4rows (in 1-12)

• 2019.11.05: Comparison to DiSTATIS

  • (1) DiSTATIS with Gorden's parcellation

  • (2) DiSTATIS with mean connectivity based on Gordon's parcellation

  • (3) DiSTATIS with mean connectivity based on individual parcellation

  • Our method: HMFA with subject first and edgetype (between vs. within) then

  • Expectation:

    • Compared to (1), (2) will miss some effect because the pattern within each edge is smoothed and the some effect will be missed.

    • Compared to (3), (2) will be less likely to detect effects in small networks

    need another simulation in small network

    • Compare to our SVD, (3) will miss the effect of patterns