Calculate group averages

using UnfoldBIDS, PyMNE
using Unfold
using DataFrames
using Statistics
using CairoMakie, AlgebraOfGraphics
using LazyArtifacts

    CondaPkg Found dependencies: /home/runner/.julia/packages/CondaPkg/8GjrP/CondaPkg.toml
    CondaPkg Found dependencies: /home/runner/.julia/packages/PyMNE/cNGDN/CondaPkg.toml
    CondaPkg Found dependencies: /home/runner/.julia/packages/PythonCall/83z4q/CondaPkg.toml
    CondaPkg Initialising pixi
             │ /home/runner/.julia/artifacts/cefba4912c2b400756d043a2563ef77a0088866b/bin/pixi
             │ init
             │ --format pixi
             └ /home/runner/work/UnfoldBIDS.jl/UnfoldBIDS.jl/docs/.CondaPkg
✔ Created /home/runner/work/UnfoldBIDS.jl/UnfoldBIDS.jl/docs/.CondaPkg/pixi.toml
    CondaPkg Wrote /home/runner/work/UnfoldBIDS.jl/UnfoldBIDS.jl/docs/.CondaPkg/pixi.toml
             │ [dependencies]
             │ openssl = ">=3, <3.6"
             │ libstdcxx = ">=3.4,<15.0"
             │ uv = ">=0.4"
             │ libstdcxx-ng = ">=3.4,<15.0"
             │
             │     [dependencies.python]
             │     channel = "conda-forge"
             │     build = "*cp*"
             │     version = ">=3.10,!=3.14.0,!=3.14.1,<4, >=3.4,<4"
             │
             │ [project]
             │ name = ".CondaPkg"
             │ platforms = ["linux-64"]
             │ channels = ["conda-forge", "anaconda"]
             │ channel-priority = "strict"
             │ description = "automatically generated by CondaPkg.jl"
             │
             │ [pypi-dependencies]
             └ mne = ">=1.4"
    CondaPkg Installing packages
             │ /home/runner/.julia/artifacts/cefba4912c2b400756d043a2563ef77a0088866b/bin/pixi
             │ install
             └ --manifest-path /home/runner/work/UnfoldBIDS.jl/UnfoldBIDS.jl/docs/.CondaPkg/pixi.toml
✔ The default environment has been installed.

Analysis

First let's redo the steps from the quickstart tutorial

sample_data_path = UnfoldBIDS.erp_core_example();
layout_df = bids_layout(sample_data_path, derivatives=false);
data_df = load_bids_eeg_data(layout_df);


33.3%┣██████████████▍                            ┫ 1/3 [00:00<Inf:Inf, InfGs/it]

66.7%┣████████████████████████████████                ┫ 2/3 [00:00<00:00, 2it/s]

100.0%┣███████████████████████████████████████████████┫ 3/3 [00:01<00:00, 4it/s]

Calculate results

basisfunction = firbasis(τ=(-0.2,.8),sfreq=1024)
f  = @formula 0~1
bfDict = ["stimulus"=>(f,basisfunction)]
UnfoldBIDS.rename_to_latency(data_df, :sample); # Unfold expects a :latency collumn in your events; if your event latency is named differently you can use this function as remedy

resultsAll = run_unfold(data_df, bfDict; eventcolumn="trial_type");


33.3%┣██████████████▍                            ┫ 1/3 [00:00<Inf:Inf, InfGs/it]

Progress:   6%|██▌                                      |  ETA: 0:00:06
Progress:  36%|██████████████▉                          |  ETA: 0:00:02
Progress:  85%|██████████████████████████████████▊      |  ETA: 0:00:00
Progress: 100%|█████████████████████████████████████████| Time: 0:00:01

66.7%┣████████████████████████████████                ┫ 2/3 [00:08<00:08, 8s/it]

Progress:  64%|██████████████████████████▏              |  ETA: 0:00:00
Progress: 100%|█████████████████████████████████████████| Time: 0:00:00

100.0%┣███████████████████████████████████████████████┫ 3/3 [00:09<00:00, 5s/it]

Progress:   6%|██▌                                      |  ETA: 0:00:09
Progress:  76%|███████████████████████████████          |  ETA: 0:00:00
Progress: 100%|█████████████████████████████████████████| Time: 0:00:00

Now, let's transform the data into a tidier format (Note: We use raw data without a high-pass filter here so estimates will be quite off)

tidy_df = unpack_results(bids_coeftable(resultsAll))
first(tidy_df, 5)

5×11 DataFrame

Row	subject	ses	task	run	channel	coefname	estimate	eventname	group	stderror	time
	SubStrin…	Missing	SubStrin…	Missing	Int64	String	Float64	String	Nothing	Nothing	Float64
1	001	missing	P3	missing	1	(Intercept)	-17604.5	stimulus			-0.200195
2	001	missing	P3	missing	2	(Intercept)	-3162.93	stimulus			-0.200195
3	001	missing	P3	missing	3	(Intercept)	11587.9	stimulus			-0.200195
4	001	missing	P3	missing	4	(Intercept)	-1056.33	stimulus			-0.200195
5	001	missing	P3	missing	5	(Intercept)	-480.149	stimulus			-0.200195

Calculate average over subjects

mean_df = combine(groupby(tidy_df, [:time, :coefname]), :estimate => mean)
first(mean_df, 5)

5×3 DataFrame

Row	time	coefname	estimate_mean
	Float64	String	Float64
1	-0.200195	(Intercept)	-5304.91
2	-0.199219	(Intercept)	-5304.87
3	-0.198242	(Intercept)	-5304.87
4	-0.197266	(Intercept)	-5304.96
5	-0.196289	(Intercept)	-5305.04

Importantly, the above can be extended to groupbyan arbitrary number of covariates!

Plot results using AOG

plt = data(mean_df) * mapping(:time, :estimate_mean, color = :coefname, group=:coefname => nonnumeric) * visual(Lines)
draw(plt, axis=(yticklabelsvisible=false,))

This page was generated using Literate.jl.