Algorithm A1: Impute missing values with random forest [28].

Require: X is an n × p matrix, setup stopping criterion ($\gamma$)   setup initial guess for missing values;   $\mathbf{k}\leftarrow$ vector of sorted indices of columns in $\mathbf{X}$ w.r.t. increasing amount of missing values;   while not $\gamma$ do    ${\mathbf{X}}_{old}^{imp}\leftarrow$ store previously imputed matrix;    for s in $\mathbf{k}$ do     Fit a random forest: ${\mathbf{y}}_{obs}^{\left(s\right)}\sim {\mathbf{x}}_{obs}^{\left(s\right)}$;     Predict ${\mathbf{y}}_{mis}^{\left(s\right)}$ using ${\mathbf{x}}_{mis}^{\left(s\right)}$;     ${\mathbf{X}}_{new}^{imp}\leftarrow$ update imputed matrix, using predicted ${\mathbf{y}}_{mis}^{\left(s\right)}$;     end for     update $\gamma$    end while    return the imputed matrix ${\mathbf{X}}^{imp}$