Suggestions

:speech_balloon:

Add a Restart-At-Random-Times option for Historical Sampling

The current “Random Start Years, Radom-Restart” option for using historical data only restarts when it gets to the end of the historic data. This means that when running too many trials, a lot of them will be totally identical. It would be nice if there was an option to restart at random times prior to reaching the end of the data. That way, it would be possible to get a larger variety of trials without totally losing the value of chronological years.

It would be probably be good to give people a way to control how often the reset happened because too many restarts could make the data less realistic.

  • Sign in to comment and vote. Sign in by email

  • 12 September 2024 Kim W suggested this task

  • 12 September 2024 Kyle Nolan approved this task

  • avatar
    Adam Golman

    This is a great idea. I ran into the same issue with identical trials.

    I was thinking of possibly two implementations.

    Option 1: User specifies the # of sequential years to “bucket” and then randomly select to generate a sequence.

    For example: N=10 –> sequence is 1930-1940, 2000-2010, 1950-1960, etc.

    Option 2: User specifies the range of #’s of sequential years and engine randomly selects from the range for sequential buckets.

    For example: N=5-10 –> sequence is 1930-1935, 2000-2009, 1950-1957, etc.

    26 September 2024
  • avatar
    Jelke van Hoorn

    I would suggest to use a distribution for the length of each streak of years. So at the start of each set of sequential years the length of this set is drawn from a distribution.

    12 February