Skip to content

Attention

The contents of this documentation might slightly differ from the information published by Abdol et al., 20211. For general introduction to SAM, please refer to the main publication while we sync the two sources.

Configuration File

SAM uses a JSON file to load and save all simulation parameters. The code block below shows a general configuration file used by SAM. As highlighted, the config file is separated into 4 different sections, each corresponding to one of SAM's component discussed previously. In this chapter, we will discuss each section, and learn how to configure each module separately; in order to be able to customize and configure all processes of execution, research and hacking workflows.

Sample Configuration File

{
    "experiment_parameters": {
        "n_reps": 1,
        "n_conditions": 2,
        "n_dep_vars": 2,
        "n_obs": 25,
        "data_strategy": {
            "name": "LinearModel",
            "measurements": {
                "dist": "mvnorm_distribution",
                "means": [0.0, 0.0, 0.4, 0.4],
                "covs": 0.5,
                "stddevs": 1.0
            }
        },
        "effect_strategy": {
            "name": "MeanDifference"
        },
        "test_strategy": {
            "name": "TTest",
            "alpha": 0.005,
            "alternative": "TwoSided",
            "var_equal": true
        }
    },
    "researcher_parameters": {
        "decision_strategy": {
            "name": "DefaultDecisionMaker",
            "initial_selection_policies": [
                ["sig", "min(pvalue)"]
            ],
            "will_start_hacking_decision_policies": [
                "!sig"
            ],
            "stashing_policy": [
                ""
            ],
            "between_hacks_selection_policies": [
                [""]
            ],
            "between_replications_selection_policies": [
                [""]
            ],
            "will_continue_replicating_decision_policy": [
                ""
            ],
            "submission_decision_policies": [
                ""
            ]
        },
        "probability_of_being_a_hacker": 1,
        "probability_of_committing_a_hack": 1,
        "hacking_strategies": [],
        "is_pre_processing": false,
        "pre_processing_methods": [
            ""
        ]
    },
    "journal_parameters": {
        "max_pubs": 24,
        "selection_strategy": {
            "name": "SignificantSelection",
            "alpha": 0.05,
            "pub_bias": 0.1,
            "side": 0
        }
    },
    "simulation_parameters": {
        "log_level": "info",
        "master_seed": "random",
        "n_sims": 1,
        "output_path": "../outputs/",
        "output_prefix": "sample_sim",
        "update_config": true,
        "progress": false,
        "save_all_pubs": false,
        "save_meta": true,
        "save_overall_summaries": true,
        "save_pubs_per_sim_summaries": true,
        "save_rejected": false
    }
}
About JSON

For the introduction to JSON, see here or here.

  1. Amir M. Abdol and Jelte M. Wicherts. Science Abstract Model Simulation Framework. PsyArXiv, 09 2021. URL: https://psyarxiv.com/zy29t, doi:10.31234/osf.io/zy29t. ↩

Last update: 2021-09-18