Tests fail when cell outputs include current timestamp

[HelenCEBM]

E.g. https://github.com/ebmdatalab/tamoxifen-survey-study-notebook/commit/b11f2f1cdcba643b3ccde2d9f2cb995e0cbdd619/checks?check_suite_id=473974680

[sebbacon]

Ideally we do want to test the output is as expected, but unfortunately the default output of the panel regression includes the current time stamp.

There are three possible approaches to this.

1. Skip checking output

• This is a last resort as we'd really like tests to do something

2. Add a regular expression here to replace timestamps in panel regression summaries with a string like xx:xx

• Doable, but because times and dates might (in other contexts) be relevant, we'd have to include panel-regression-specific identifiers in the the regex, and adding code for one odd exception isn't great in terms of long term framework maintenance (the current regexes linked above are globally and safely applicable)

3. Change the output not to include time and date stamps

• The current output is the default representation of a PanelEffectsResults which I presume is summary
• The fix would be to specifically output the results we're interested in, rather than the kitchen sink (e.g. rsquared, pvalues etc)
  • However, if we are interested in basically everything except the date + time the regression was run, this will be a bit arduous

Thoughts?

[CarolineMorton]

Option 3: We could probably get away with the Parameter Estimates, Log Likelihood, Entities I suspect. Could we add in some magic that allows us to pass in arguments to keep (in this case the above) so that it can be reused for other types of panels?

[madwort]

Option 4: use freezegun to ensure that the current time stamp is always some preset value