Calibration for a time-dependent variable - any sugestions?
Marcio Naves
Hello everyone,
I am modelling a Markov cohort for cost-utility analysis, that includes two time-dependent variables (death and hospitalization) for which I created tables to account for age progression.
So my question is: how to calibrate these age-dependent probabilities? Apparentely, I can only input individual parameters when using calibration.
Comments
Yes, it is true that TreeAge Pro's calibration option only adjusts inputs variables when attempting to match targets. Even theoretically, I don't think it would be possible to independently adjust a large set of time-dependent inputs on a limited set of targets.
If you are prepared to accept your assumption of the relationship among the time-dependent values, you could adjust the full set of values up and down using a single variable. Then you could use that single variable as in input in the calibration process.
Note that there is a feature available that might help you with this. From the menu, you can choose Tree < Expand Table for Sensitivity Analysis. This will expand the table into 3 columns for low, base case and high values. While this is designed for sensitivity analysis, you could use the new value column variable for calibration as well.
Andrew
Thank you Andrew, for your quick reply and considerations. The new option to expand the table for sensitivity analysis will work well for my model.
Marcio
Happy to hear it.
Andrew
Please sign in to leave a comment.