ALE parameters

Configuration Parameters for ALE

This page provides a detailed explanation of the parameters available in the config.yml file. These settings control various aspects of the ALE analysis workflow. Please note: Many of these parameters are intended for advanced users. DO NOT CHANGE THESE IF YOU ARE NOT AN ALE EXPERT!

---

Project Folder

The Project Folder section specifies the file paths for key project documents:

• analysis_info

  • Default: "analysis_info.xlsx"
  • Description: Excel file containing analysis-related information.

• experiment_info

  • Default: "experiment_info.xlsx"
  • Description: Excel file containing details about the experiments.

---

ALE Parameters

These parameters control the main ALE analysis settings:

• pool_experiments
  • Default: False
  • Description: When enabled, multiple experiments from the same paper are pooled into a single experiment. This helps in cases where individual experiments are not statistically independent. Please note, pooling only makes sense when doing an analysis across all experiments. For example, in a meta-analysis on emotion processing, a study can contribute multiple experiments by reporting coordinates for positive but also negative stimuli. In case you have coded the valence in your experiment_info file and run an analysis across all experiments inluding positive and negative stimuli, studies reporting coordinates of both positive and negative stimuli will be treated as two experiments although they come from the same participant group. Here, enabling pooling_experiments would be useful. However, if two separate analyses with one for positive and another for negative stimuli should be calculated, enabling pooling should not be done.

Please be aware that two rows of the experiment_info file are only treated as the same experiment, when all column values of the two rows are identical. Thus, it is often helpful to only include columns in your experiment_info file, which are relevant for the analysis. For example, when running two analyses for positive and negative stimuli, only keep the columns of "Article", "Subjects", "x", "y", "z", "space" and the column that codes the "valence" in the file and remove potential other tags (for example an additional column with information on the specific type of valence, i.e. angry, fearful) that could lead to separation of studies into different experiments.

Often it is therefore necessary to calculate different analyses by using different experiment_info and analysis files.

• tfce_enabled

  • Default: True
  • Description: Enables TFCE (Threshold-Free Cluster Enhancement) for multiple comparison correction. (See Frahm et al., 2022 for details.)

• gm_masking

  • Default: True
  • Description: If enabled, the ALE map—and all subsequent maps—will be masked by the ICBM 10% GM mask to restrict the analysis to gray matter regions.

• bin_steps

  • Default: 0.0001
  • Description: Defines the size of the bins used in the Modeled Activation (MA) histogram. Smaller steps provide a finer resolution of the histogram.

• cutoff_predict_enabled

  • Default: True
  • Description: When enabled, xgboost models are used to predict cutoff values instead of relying on monte carlo simulation. (Refer to Frahm et al., 2024.)

• significance_threshold

  • Default: 0.05
  • Description: The p-value required for significance. Important: This setting only applies if cutoff_predict_enabled is disabled.

• cluster_forming_threshold

  • Default: 0.001
  • Description: Sets the preliminary cluster forming threshold for cluster-level family-wise error correction. Important: This is only effective if cutoff_predict_enabled is disabled.

• monte_carlo_iterations

  • Default: 5000
  • Description: Number of iterations used for monte-carlo based multiple comparison correction. Note: Only applicable when cutoff_predict_enabled is disabled.

• subsample_n

  • Default: 2500
  • Description: The number of subsamples calculated for the probabilistic ALE algorithm.

• contrast_permutations

  • Default: 10000
  • Description: Specifies the number of iterations used in the classic contrast algorithm.

• contrast_correction_method

  • Default: "cFWE"
  • Options: "cFWE", "vFWE", "tfce"
  • Description: Determines the correction method for the contrast algorithm.
    • "cFWE": Cluster-level family-wise error correction (default).
    • "vFWE": Voxel-level family-wise error correction.
    • "tfce": Threshold-Free Cluster Enhancement correction.

• difference_iterations

  • Default: 1000
  • Description: Number of sub-iterations used in the balanced contrast algorithm.

• nprocesses

  • Default: 2
  • Description: Sets the number of parallel processes for various steps in the ALE algorithm. Adjust this number based on the capabilities of your machine.

---

MA_Clustering Parameters

These parameters control the settings for the Modeled Activation (MA) clustering process. DO NOT CHANGE THESE IF YOU ARE NOT AN ALE EXPERT!

• max_clusters

  • Default: 10
  • Description: The maximum number of clusters allowed in the analysis.

• subsample_fraction

  • Default: 0.9
  • Description: The fraction of data used in each subsampling iteration during the clustering process.

• sampling_iterations

  • Default: 1000
  • Description: Number of iterations for subsampling in the clustering algorithm.

• null_iterations

  • Default: 5000
  • Description: Number of iterations used to generate the null distribution for statistical testing.

• correlation_type

  • Default: "spearman"
  • Options: "spearman", "pearson"
  • Description: Specifies the type of correlation coefficient used to assess similarity between clusters.

• clustering_method

  • Default: "hierarchical"
  • Options: "hierarchical", "kmeans"
  • Description: Determines the clustering method. Hierarchical clustering is the default.

• linkage_method

  • Default: "complete"
  • Options: "complete", "average", "ward"
  • Description: Defines the linkage method for hierarchical clustering.

---

Important Notes

• Modifying Parameters: Changes to the config.yml file can significantly affect the outcome of your analyses. Only adjust these parameters if you fully understand their implications.
• Further Reading: For more detailed explanations and the latest updates, please refer to the ALE documentation and the cited publications (e.g., Frahm et al., 2022; Frahm et al., 2024).