Group-Level Statistical Analysis¶

SPIMquant supports group-level statistical analysis to compare pathology metrics across experimental groups.

Overview¶

Group analysis performs statistical comparisons of:

Field fraction: Proportion of pathology signal
Density: Count-based metrics
Volume: Volumetric measurements
Custom metrics: User-defined quantifications

Results are provided as:

Statistical tables (t-statistics, p-values, effect sizes)
Heatmap visualizations
3D volumetric maps
Group-averaged statistics

Prerequisites¶

Before running group analysis:

Complete participant-level processing for all subjects
Create participants.tsv with group assignments
Define contrast column and values

Creating participants.tsv¶

Create a TSV file in your BIDS directory with participant metadata:

participant_id  treatment   age sex genotype
sub-01  control 12  M   WT
sub-02  control 13  F   WT
sub-03  drug    11  M   WT
sub-04  drug    12  F   WT
sub-05  control 12  M   KO
sub-06  drug    11  M   KO

Required columns:

participant_id: Subject identifiers (must match BIDS subject IDs)
At least one grouping column (e.g., treatment, genotype)

Running Group Analysis¶

Basic Group Comparison¶

Compare two groups:

pixi run spimquant /bids /output group \
  --contrast_column treatment \
  --contrast_values control drug \
  --cores all

This compares:

Group 1: treatment == control
Group 2: treatment == drug

Multiple Contrasts¶

# Compare multiple grouping variables
pixi run spimquant /bids /output group \
  --contrast_column genotype \
  --contrast_values WT KO \
  --cores all

Output Files¶

Group analysis generates several output files:

Statistical Results¶

*_groupstats.tsv: Statistical test results by region

region  t_statistic p_value effect_size mean_control    mean_drug
ctx-lh-frontal  3.45    0.023   0.82    0.145   0.234
ctx-lh-temporal 2.11    0.058   0.51    0.098   0.143
...

Columns:

region: Brain region name
t_statistic: t-test statistic
p_value: Uncorrected p-value
effect_size: Cohen's d effect size
mean_<group>: Group mean values

Visualizations¶

*_groupstats.png: Heatmap of statistical results

Shows color-coded significance across brain regions.

3D Maps¶

*_groupstats.nii: Volumetric maps of statistics

Contains:

t-statistic maps
p-value maps
Effect size maps

View in neuroimaging software (FSLeyes, ITK-SNAP, 3D Slicer).

Group Averages¶

*_groupavgsegstats.tsv: Average statistics per group

region  group   mean_fieldfrac  std_fieldfrac   n_subjects
ctx-lh-frontal  control 0.145   0.023   10
ctx-lh-frontal  drug    0.234   0.034   10
...

*_groupavg.nii.gz: Group-averaged volumetric maps

Separate volumes for each group showing average pathology distribution.

Statistical Methods¶

SPIMquant uses:

t-tests: For two-group comparisons
Multiple comparison correction: Optional FDR/Bonferroni
Effect size: Cohen's d for interpretability

Filtering and Quality Control¶

Excluding Subjects¶

Exclude specific subjects from analysis:

pixi run spimquant ... group \
  --exclude_subjects 05 07 \
  --contrast_column treatment \
  --contrast_values control drug

Region Filtering¶

Focus analysis on specific brain regions.

Visualization Options¶

Customize Heatmaps¶

Control heatmap appearance:

# TODO: Add heatmap customization options

Export for External Tools¶

Results can be imported into:

R: For custom statistical analysis
Python: Using pandas/seaborn
GraphPad Prism: For publication figures

Advanced Analysis¶

Covariates¶

Include covariates in statistical models:

# TODO: Add covariate examples

Multiple Testing Correction¶

Apply FDR or Bonferroni correction:

# TODO: Add correction examples

Custom Contrasts¶

Define complex contrasts:

# TODO: Add custom contrast examples

Example Workflows¶

Drug Treatment Study¶

# 1. Process all subjects
pixi run spimquant /bids /output participant --cores all

# 2. Run group analysis
pixi run spimquant /bids /output group \
  --contrast_column treatment \
  --contrast_values vehicle drug \
  --cores all

Genotype Comparison¶

# Compare wildtype vs knockout
pixi run spimquant /bids /output group \
  --contrast_column genotype \
  --contrast_values WT KO \
  --cores all

Multi-Factor Design¶

# TODO: Add interaction effects

Interpreting Results¶

Statistical Significance¶

p < 0.05: Traditionally significant
p < 0.01: Highly significant
Consider effect sizes alongside p-values

Effect Sizes¶

Cohen's d interpretation:

d < 0.2: Small effect
0.2 ≤ d < 0.8: Medium effect
d ≥ 0.8: Large effect

Visualization¶

Red/warm colors: Higher in group 2
Blue/cool colors: Higher in group 1
Intensity: Magnitude of difference

Troubleshooting¶

Missing participants.tsv¶

Error: participants.tsv not found

Create participants.tsv in BIDS root directory
Ensure TSV format (tab-separated)

Group Size Warnings¶

Warnings about small group sizes:

Minimum 3 subjects per group recommended
Larger groups increase statistical power

Incomplete Data¶

Some subjects missing outputs:

Ensure all subjects completed participant-level
Check for failed jobs in participant-level logs

Next Steps¶

Visualization: Advanced visualization techniques
Examples: Complete analysis examples
FAQ: Common questions about group analysis