Skip to content

Version 1.3 Updates

Several updates have been made to address bugs, improve the robustness of the estimation procedure, and add features.

Backward Compatibility Issues

Version 1.3 introduces a change that impacts compatibility with version 1.2. There has been a change to the format of omit_fixed_effect argument of EstimationModel such that past usage will generally not be compatible in version 1.3. See the EstimationModel documentation for details on the new format.

Bug Fixes

  • Fixed error in format_regression_table in which the default significance levels were 0.1, 0.05, and 0.1 rather than 0.01, 0.05, and 0.1. The error was causing estimates to receive *** when they were only 10% significant.

  • Fixed issue with the retention of record identifiers in EstimationModel.modified_data after estimation. Previously, the modified estimation data did not include identifiers like importer, exporter, or year. These identifiers are now included in the modified data returned after estimation.

  • Fixed issue in which pre-estimation diagnostics would drop observations associated with collinear fixed effects if the trade values of those observations were all greater than zero.

Improved Robustness

  • The omit_fixed_effect argument has been made more robust. It now allows for users to select specific fixed effects to drop in order to avoid collinearity. Notably, the new implementation uses dictionary inputs instead of list inputs so that older uses will no longer work properly in version 1.3. See

New Features

  • Option for clustered standard errors (cluster_on) has been added to EstimationModel.

  • Added method for creating concatenated 'pair' variables from categorical variables to EstimationData (add_pair_var).

  • Added make_data_square function.

  • Added correlation method to EstimationData, which produces a correlation table and visualization.

  • Added kernel density plot function coefficient_kd_plot to visualize sector-by-sector estimated results.

  • Tables created by format_regression_table now report the significance levels at the bottom of the table when written to a file. There is now also an optional notes argument to add a user-supplied note to the bottom of the written table.