Default parameters

[gaow]

Required: X, Y

Model size / convergence:

• L = 20
• L_greedy = 5
• greedy_lbf_cutoff = 0.1
• max_iter = 50 (was 100)
• tol = 1e-4
• convergence_method = "pip"
• pip_stall_window = 5L

Prior:

• prior_variance_grid = NULL
• prior_variance_scope = "per_outcome" # new one to try is "per_scale_normal" using a simple point-normal prior
• null_prior_init = 0
• cross_outcome_prior = NULL
• prior_weights = NULL
• estimate_prior_variance = TRUE

Residual variance:

• residual_variance = NULL
• residual_variance_scope = "per_outcome"
• estimate_residual_variance = TRUE

X preprocessing: standardize = TRUE, intercept = TRUE

CS: coverage = 0.95, min_abs_corr = 0.5

Wavelet pipeline:

• pos = NULL
• max_padded_log2 = 10
• wavelet_basis_order = 10 (renamed from wavelet_filter_number)
• wavelet_family = "DaubLeAsymm"
• wavelet_magnitude_cutoff = 0 (renamed from low_count_filter)
• wavelet_qnorm = FALSE (renamed from quantile_norm)
• wavelet_standardize = TRUE (added to interface where original fsusieR hard coded)

mixsqp:

• control_mixsqp = NULL,
• mixture_null_weight = 0.05 (previously 0.1)

mixsqp and ebnm:

• alpha_thin_eps = 5e-5 (to consider all variants with single effect PIP > 5E-5 instead of just top 100)

Other:

• small_sample_correction = FALSE
• model_init = NULL
• verbose = TRUE, track_fit = FALSE
• attach_smoothing_inputs = TRUE