** TODOs

• select absolute weight threshold for edges.
• show differences.
• select/hilite taxa based on properties like sparsity. ** 10 July 2024 [[file:data-qsip-jeff/figure-scatter/]] directory has analysis of 7 SOAK experiments.

[[file:data-qsip-jeff/figure-scatter/figure-scatter-print.R]] makes

[[file:data-qsip-jeff/figure-scatter/figure-scatter-print.png]]

[[file:data-qsip-jeff/figure-scatter/figure-scatter-interactive.R]] makes https://tdhock.github.io/2024-07-10-error-predict-qsip/ (screenshot below)

[[file:data-qsip-jeff/figure-scatter/figure-scatter-interactive.png]]

** 12 Jan 2023

[[file:HOCKING-slides-2024-01-12-berkeley.tex]] makes [[file:HOCKING-slides-2024-01-12-berkeley.pdf]] for presentation, based on qsip_pc2_all_new- and drawing-cv-same-other figures.

** 8 Jan 2023

[[file:data-qsip-jeff/qsip_pc2_all_new-analyze.R]] reads RData files and creates

• 6 test groups: [[file:data-qsip-jeff/qsip_pc2_all_new-controls.between.sites.png]]
• 3 test groups: [[file:data-qsip-jeff/qsip_pc2_all_new-control.vs.carbon.additions.png]]
• 2 test groups: [[file:data-qsip-jeff/qsip_pc2_all_new-controls.between.experiments.png]]

+BEGIN_SRC R

comparison.dt comparison group.var DT 1: controls.between.sites site <data.table[17225x8407]> 2: control.vs.carbon.additions treatment <data.table[60877x8407]> 3: controls.between.experiments experiment <data.table[7710x8407]> length(gene.names <- grep("^K", names(qsip.dt), value=TRUE)) [1] 8380 for(comparison.i in 1:nrow(comparison.dt)){

• compare.row <- comparison.dt[comparison.i]
• print(compare.row$DT[[1]][, .(rows=.N), by=eval(compare.row$group.var)])
• } site rows 1: glacier_forefield 767 2: litchfield_island 1043 3: GL 3560 4: MC 3120 5: PJ 4317 6: PP 4418 treatment rows 1: control 17225 2: C 23214 3: CN 20438 experiment rows 1: dim 3120 2: qme 4590

  +END_SRC

** 5 Jan 2023

[[file:data-2023-12-22-benchmark.R]] makes [[file:data-2023-12-22-benchmark.RData]] with same/other experiment on necromass data. Next [[file:data-2023-12-22-benchmark-analyze.R]] makes [[file:data-2023-12-22-benchmark-analyze.png]] which visualizes the prediction error for every taxon and group. cv_glmnet seems to be better than featureless for most taxa, at least for train.groups=all.

[[file:data-qsip-jeff/qsip_pc2_all_new.R]] runs same/other experiments on large qsip data set, making

• [[file:data-qsip-jeff/qsip_pc2_all_new-controls.between.experiments.RData]]
• and two more.

[[file:data-qsip-linnea/qSIP_rates_tax.R]] loads Steve/Linneas data, not sure what to do with this, wrote email, waiting for response.

** 10 Oct 2023

[[file:figure-one-network.R]] visualizes network/scatter data as below, (click image below to render interactive viz)

[[https://tdhock.github.io/necromass-figure-one-network][file:figure-one-network.png]]