Closed FTamas77 closed 1 month ago
Hi @FTamas77 ,
The m
is an array that we are creating to draw the true DAG, so we do not use it to run RESIT.
Like the inputs and outputs of other algorithms, the dataset is the input and the adjacency matrix is the output. However, because of RESIT's assumption of nonlinear functions, the adjacency matrix will have a value of 0 or 1.
@ikeuchi-screen Thank you very much for the clarification. So, is it correct if I say that one means an existing correlation and zero means no correlation between the variables (in the adjacency matrix that is calculated during the fitting)?
And I guess the model.bootstrap(X, n_sampling=n_sampling)
is good if I want to quantify this causal effect, isn't it?
@FTamas77 One in the adjacency matrix represents an edge in the causal graph, while zero represents no edge. Edges denote causation rather than correlation.
Because the relationship between variables is nonlinear, even model.bootstrap
cannot estimate causal effects. Bootstrapping can estimate the probability of edges.
Thank you very much for the clarification.
For example, we have this code here: https://github.com/cdt15/lingam/blob/master/docs/tutorial/resit.rst
In that, we create an array:
m = np.array([ [0, 0, 0, 0, 0], [1, 0, 0, 0, 0], [1, 1, 0, 0, 0], [0, 1, 1, 0, 0], [0, 0, 0, 1, 0]])
But we never use it. I am confused because the adjacency matrix contains only 1 after the fitting. Could you help me to clarify the input/output of that code?