Skip to content

Configuration Parameters

This page describes the parameters in setparams.m. This file is located in the project root directory and serves as the configuration file for PRISME. Please modify it to run your analysis. We recommend reviewing the file before going through this guide.

Quick Reference

  • data_dir - Path to your dataset used for the analysis
  • list_of_nsubset - List of sample sizes to perform the power calculation
  • n_repetitions - Number of subsampling repetitions to be used
  • all_cluster_stat_types - List of statistical inference methods used in the analysis
  • n_perms - Number of permutations used for the non-parametric inference methods

Data and Output Paths

All attributes here are fields in the Params struct defined in the setparams.m file.

data_dir (string, required)

Path to the input dataset .mat file used for the power calculation. See Input Data Format for more instructions.

Params.data_dir = './data/my_dataset.mat';

output (string, optional)

Name for the output directory. If not specified, defaults to the dataset filename. PRISME saves the results from the subsampling repetitions in a directory folder called repetitions, the ground truth calculation in ground_truth, and the power calculation in power_calculation. All three folders are contained in a directory named after Params.output within power_calculator_results.

Params.output = 'my_power_analysis';

save_directory (string, optional)

The directory where all results are saved. Default: './power_calculator_results/'

Params.save_directory = './my_results/';

subsample_file_type (string, optional)

Output file format: 'full_file' or 'compact_file'. Full files save all relevant p-values and all estimated t-statistics. Therefore, they consume significantly more storage and are slower to checkpoint. Compact files only save the average t-statistic and the number of times each variable was found significant. Default: 'compact_file'

Params.subsample_file_type = 'compact_file';

atlas_file (string or NaN, optional)

Path to a custom brain atlas file. Default: NaN. If no atlas is provided, PRISME cannot assign variables to networks, and network-level inference methods will not be performed.

Params.atlas_file = './atlases/schaefer_268.mat';
% or
Params.atlas_file = NaN;  % No atlas

recalculate (boolean, optional)

If true, recalculates existing results. If false, resumes from existing checkpoints. Default: false

Params.recalculate = true;  % Force recalculation

list_of_nsubset (cell array, required)

Sample sizes to test in your power analysis.

Params.list_of_nsubset = {20, 40, 80, 120, 200};

Note: For two-sample t-tests, this is the size per group (total N = 2n).

n_repetitions (integer, required)

Number of subsampling repetitions.

Params.n_repetitions = 100;

batch_size (integer, optional)

Number of repetitions per checkpoint batch. Results are only saved to disk once a batch is completed. Therefore, this parameter balances how much data will be stored in RAM before saving it to disk and checkpointing. The larger it is, the faster the overall calculation, but the more RAM memory will be necessary.

Params.batch_size = 50;  % Process 50 repetitions before saving

tests_to_skip (function handle, optional)

Function handle that defines which studies to skip based on their index. Default skips nothing. This will be replaced in a future update.

% Skip studies 5-10
ranges = {[5, 10]};
Params.tests_to_skip = @(x) any(cellfun(@(r) (x >= r(1)) && (x <= r(2)), ranges));

Statistical Settings

all_cluster_stat_types (cell array, required)

Statistical inference methods to evaluate. Some of the available methods:

  • 'Parametric' - Parametric inference using t-distribution
  • 'Size_cpp' - Cluster-size inference with C++ implementation
  • 'Fast_TFCE_cpp' - TFCE with incremental cluster and C++
  • 'Constrained_cpp' - cNBS with C++ implementation
  • 'Omnibus_cNBS' - Multivariate whole-brain test

Non-C++ versions are named identically without the _cpp suffix.

Params.all_cluster_stat_types = {'Parametric', 'Fast_TFCE_cpp', 'Constrained_cpp'};

all_submethods (cell array, optional)

Multiple comparison correction methods. Current options: 'FWER', 'FDR'

Params.all_submethods = {'FWER', 'FDR'};

n_perms (integer, required for non-parametric methods)

Number of permutations for non-parametric inference.

Params.n_perms = 1000;

force_permute (boolean, optional)

If true, forces permutation generation even if parametric methods that don't require permutations are chosen. Default: false

Params.force_permute = true;

pthresh_second_level (float, optional)

Significance threshold (alpha level). This significance threshold is applied to multiple-comparison corrected p-values when applicable.

Params.pthresh_second_level = 0.05;  % 5% significance level

tthresh_first_level (float, optional)

T-statistic threshold for initial cluster formation. Only used by the Cluster Size method. Default: 3.1. This parameter will be moved into method-specific implementations in a future update.

Params.tthresh_first_level = 3.1;  % Approximately p=0.001-0.005

tpr_dthresh (float, optional)

Threshold for defining true effects from ground truth. If the magnitude is above this value, the ground truth effect is considered a true effect. Default: 0

Params.tpr_dthresh = 0;

save_significance_thresh (float, optional)

Only p-values below this threshold are saved to disk (memory optimisation). Default: 0.15. This only applies to the full_file structure. Compact files only save the number of times a variable was found significant.

Params.save_significance_thresh = 0.15;

cluster_size_type (string, optional)

For Cluster Size method: 'Extent' (cluster size) or 'Intensity' (cluster mass). Default: 'Extent'. This parameter will be moved into method-specific implementations in a future update.

Params.cluster_size_type = 'Extent';

Parallel Processing

parallel (boolean, optional)

Enable parallel processing across repetitions. Default: true

Params.parallel = true;

n_workers (integer, optional)

Number of parallel workers. Set to the number of available CPU cores. Default: 10

Params.n_workers = 16;  % Use 16 cores

Advanced Parameters

gt_origin (string, deprecated)

Ground truth calculation origin. Deprecated.

Params.gt_origin = 'power_calculator';

nbs_dir (string, internal)

Path to NBS directory for internal dependencies.

Params.nbs_dir = './NBS1.2';

other_scripts_dir (string, internal)

Deprecated - path to old cNBS scripts.

Params.other_scripts_dir = './NBS_benchmarking/support_scripts/';

Developer Options

These parameters enable faster execution for testing and development. Do not use for actual power calculations.

testing (boolean, optional)

Enables testing mode with reduced parameters. Default: false

Params.testing = true;

test_n_perms (integer, optional)

Reduced permutation count for testing. Default: 10

Params.test_n_perms = 10;

test_n_repetitions (integer, optional)

Reduced repetition count for testing. Default: 5

Params.test_n_repetitions = 5;

test_n_workers (integer, optional)

Worker count for testing. Default: 1

Params.test_n_workers = 1;

test_disable_save (boolean, optional)

Disables saving results during testing. Default: false

Params.test_disable_save = true;