Call Loop and differential loop via one-way anova

[lhqing]

• [x] Call loop for L4Region
• [x] Prepare CoolDS
• [x] Aggregate all loop pixels
• [x] Make L4Region-by-loop pixel matrix
• [x] Check results

# code
ecker-hanqing-analysis/221007-combine-m3c-loops

# LoopDS, Loop-by-L4Region-by-ValueType, ValueType include E, E2, T, T2, Q, Q2, in order to run ANOVA
ecker-hanqing-wmb-us-west1/CEMBA_3C/loop/CEMBA.snm3C.LoopValues_total.ds

# Overall Loop and Loop Summit pixel sum, each L4Region is 1, the matrix value range is [0, n_l4_region]
ecker-hanqing-wmb-us-west1/CEMBA_3C/loop/CEMBA.snm3C.LoopAndSummit.ds

# Loop ANOVA stats and raw P-value
Used both Q and T; at L2 and L4Region level
ecker-hanqing-wmb-us-west1/CEMBA_3C/loop/CEMBA.snm3C.LoopANOVAStats.ds

[lhqing]

L2 L4Region level ANOVA F-stats are highly correlated, L4Region level stats are smaller so less sig, L2 level stats are larger and almost all sig.

The F-stats can be used as a way to prioritize loop cell-type-specificity, but the significance is heavily depend on what cluster level

[lhqing]

• [x] Check the loop blacklist issue, too many bins removed by blacklist
• [x] Redo loop anova using whole matrix instead of loop bins