Update pandas dependency range

[csala]

Copulas currently does not support pandas>=1.1.5, which does not include the newest versions.

We should investigate why blocks us from using the latest versions and adapt the code if necessary.

[fealho]

Currently, if we run pytest with pandas > 1.1.5 three test cases fail. After some investigation, I believe they fail because of this update on their side. This change modifies the behavior of the corr() function which we use here:

Copulas/tests/unit/multivariate/test_tree.py

Line 323 in 8d7f4e3

self.tau_mat = self.data.corr(method='kendall').values

We then use the tau matrix in the fitting process:

Copulas/tests/unit/multivariate/test_tree.py

Line 333 in 8d7f4e3

self.tree.fit(0, 4, self.tau_mat, self.u_matrix)

Since the pandas update changes our tau matrix, which in turn changes the fitting procedure, we end up generating a different tree in the end. Since the test cases expect the tree generated by the old pandas, they fail the current test.

In summary, after we correct the test cases that fail, I believe there will be no problems in updating pandas for all our libraries.

[rivendellx]

For my other project, I must use pandas 0.25.4. just wondering would it cause any issues in the correlation/covariance/tau_mat matrix produced by the newest copulas package?