Table Distribution

I'm creating a table distribution for a range of costs and utilities for a PSA. There is no sequence to the values in the table so I created the table with the index just sequential numbering. I want treeage to randomly select a value from the table in the PSA. Is it possible to set an index for this?

Thank you!

0

Comments

5 comments
  • Official comment

    Table distributions are typically used to assign characteristics to individuals in a Microsimulation based on the demographics of the cohort. However, it may be fine to use in this context as well.

    Table distributions return specific values you provide with the likelihood you provide. For example, you could want the cost to be...

    • 1000 25% of the time
    • 1500 50% of the time
    • 2000 25% of the time

    To achieve this, you would first create at table with the following data.

    • Index, Value
    • 1000, 0.25
    • 1500, 0.50
    • 2000, 0.25

    There would be no index 1, 2, 3 in a table used for this purpose.

    If you then created a table distribution using that table, it would return the values based on the likelihood provided in the first set of bullet points.

    In the typical patient characteristics context, the sampling rate for the table would be "by trial".
    To use this in your PSA context, the sampling rate would be "by EV".

    Andrew

  • That worked thank you!

  • Is it possible to use a distribution to define the likelihood of a value? Or is there an easier way to do this: 

    Gender =

    Value, Likelihood

    0, betadistribution

    1, 1-betadistribution

    Thanks

  • No, David.

    Table distributions need to have fixed values. However, you can use the beta distribution along with a uniform distribution (random from 0-1, real values) to derive a value in a different way.

    • GenderDist is a beta distribution for probability of Gender value 1 (that could be male or female).
    • GenderCheck is a uniform distribution.
    • Gender is a variable with definition Gender = if(GenderCheck <= GenderDist; 1; 2)

    This would allow you to assign the Gender characteristic based on a parameter/EV-level PSA distribution for the probability of one gender value or the other.

    Andrew

  • That makes sense - Thanks, Andrew

Please sign in to leave a comment.

Didn't find what you were looking for?

New post