FAQ: Helpful bug reporting and quick solutions

[bashtage]

The quality of a bug report directly affects the time that the bug is expected to remain open.

An ideal bug report should have two elements:

• Important: An initial description of the bug
• Essential: a copy-and-pastable demonstration, that is as small as possible, that reproduces the bug. It should:
  • Not depend on external data. While it may be possible to include an external data file (e.g., a csv), it is much simpler for developers to copy and paste code to reproduce the bug without first downloading data.
  • Use simulated data that resembles the original dataset in shape, dtype, and in any other dimensions.

Three examples that range from ideal to difficult to use:

• FAQ: get_prediction fails with regression models after calling remove_data #6887 is an excellent example of a bug report
• RegressionResultsWrapper.fvalue, typically scalar, becomes 2D array with robust standard errors #6778 is a passable example. While the code at the top is enough to reproduce the issue, it cannot be copied, pasted and executed without editing.
• VARResults.summary doesn't show all Variables #6203 is a difficult to use bug report. While it contains a skeleton of an example, it is missing key elements like the definition of the data.

Please follow this guidance when filing a report.

[everlast7]

Hello, huge fan of your work.
do you have any working examples of forecasting STLForecast function? i have tried an example, and failed to get it to go.
more importantly, would be good to see and example of prediction/forecast of the seasonal component generated by the function.

if my question does not belong here - please delete. thank you.

[bashtage]

@everlast7 Can you open a separate thread and post some example code (even if it doesn't work)?